Merge pull request #2569 from shanemcd/devel

Bump version to 2.1.0

Reviewed-by: https://github.com/softwarefactory-project-zuul[bot]

.gitignore

@@ -1,3 +1,4 @@
+
 # Tags
 .tags
 .tags1
@@ -52,6 +53,7 @@ __pycache__
 **/node_modules/**
 /tmp
 **/npm-debug.log*
+**/package-lock.json
 
 # UI build flag files
 awx/ui/.deps_built

CONTRIBUTING.md

@@ -86,8 +86,8 @@ If you're not using Docker for Mac, or Docker for Windows, you may need, or choo
 
 The AWX UI requires the following:
 
-- Node 6.x LTS version
-- NPM 3.x LTS
+- Node 8.x LTS
+- NPM 6.x LTS
 
 ### Build the environment
 
@@ -145,7 +145,7 @@ Start the development containers by running the following:
 (host)$ make docker-compose
 ```
     
-The above utilizes the image built in the previous step, and will automatically start all required services and dependent containers. Once the containers launch, your session will be attached to the *awx* container, and you'll be able to watch log messages and events in real time. You will see messages from Django, celery, and the front end build process.
+The above utilizes the image built in the previous step, and will automatically start all required services and dependent containers. Once the containers launch, your session will be attached to the *awx* container, and you'll be able to watch log messages and events in real time. You will see messages from Django and the front end build process.
 
 If you start a second terminal session, you can take a look at the running containers using the `docker ps` command. For example:
 
@@ -174,7 +174,7 @@ The first time you start the environment, database migrations need to run in ord
 ```bash
 awx_1        | Operations to perform:
 awx_1        |   Synchronize unmigrated apps: solo, api, staticfiles, debug_toolbar, messages, channels, django_extensions, ui, rest_framework, polymorphic
-awx_1        |   Apply all migrations: sso, taggit, sessions, djcelery, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
+awx_1        |   Apply all migrations: sso, taggit, sessions, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
 awx_1        | Synchronizing apps without migrations:
 awx_1        |   Creating tables...
 awx_1        |     Running deferred SQL...
@@ -329,7 +329,7 @@ We like to keep our commit history clean, and will require resubmission of pull
 
 Sometimes it might take us a while to fully review your PR. We try to keep the `devel` branch in good working order, and so we review requests carefully. Please be patient.
 
-All submitted PRs will have the linter and unit tests run against them, and the status reported in the PR.
+All submitted PRs will have the linter and unit tests run against them via Zuul, and the status reported in the PR.
 
 ## Reporting Issues
 

DATA_MIGRATION.md

@@ -38,7 +38,7 @@ Export all objects
 
 Clean up remnants of the old AWX install:
 
-```docker rm -f $(ps -aq)```     # remove all old awx containers
+```docker rm -f $(docker ps -aq)```     # remove all old awx containers
 
 ```make clean-ui```              # clean up ui artifacts
 

INSTALL.md

@@ -119,12 +119,12 @@ To complete a deployment to OpenShift, you will obviously need access to an Open
 
 You will also need to have the `oc` command in your PATH. The `install.yml` playbook will call out to `oc` when logging into, and creating objects on the cluster.
 
-The default resource requests per-pod requires:
+The default resource requests per-deployment requires:
 
 > Memory: 6GB
 > CPU: 3 cores
 
-This can be tuned by overriding the variables found in [/installer/openshift/defaults/main.yml](/installer/openshift/defaults/main.yml). Special care should be taken when doing this as undersized instances will experience crashes and resource exhaustion.
+This can be tuned by overriding the variables found in [/installer/roles/kubernetes/defaults/main.yml](/installer/roles/kubernetes/defaults/main.yml). Special care should be taken when doing this as undersized instances will experience crashes and resource exhaustion.
 
 For more detail on how resource requests are formed see: [https://docs.openshift.com/container-platform/latest/dev_guide/compute_resources.html#dev-compute-resources](https://docs.openshift.com/container-platform/latest/dev_guide/compute_resources.html#dev-compute-resources)
 
@@ -236,7 +236,7 @@ Using /etc/ansible/ansible.cfg as config file
 }
 Operations to perform:
   Synchronize unmigrated apps: solo, api, staticfiles, messages, channels, django_extensions, ui, rest_framework, polymorphic
-  Apply all migrations: sso, taggit, sessions, djcelery, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
+  Apply all migrations: sso, taggit, sessions, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
 Synchronizing apps without migrations:
   Creating tables...
     Running deferred SQL...
@@ -426,7 +426,7 @@ If you choose to use the official images then the remote host will be the one to
 
 > As mentioned above, in [Prerequisites](#prerequisites-1), the prerequisites are required on the remote host.
 
-> When deploying to a remote host, the playook does not execute tasks with the `become` option. For this reason, make sure the user that connects to the remote host has privileges to run the `docker` command. This typically means that non-privileged users need to be part of the `docker` group.
+> When deploying to a remote host, the playbook does not execute tasks with the `become` option. For this reason, make sure the user that connects to the remote host has privileges to run the `docker` command. This typically means that non-privileged users need to be part of the `docker` group.
 
 
 #### Inventory variables
@@ -449,6 +449,10 @@ Before starting the build process, review the [inventory](./installer/inventory)
 
 When using docker-compose, the `docker-compose.yml` file will be created there (default `/var/lib/awx`).
 
+*ca_trust_dir*
+
+> If you're using a non trusted CA, provide a path where the untrusted Certs are stored on your Host.
+
 #### Docker registry
 
 If you wish to tag and push built images to a Docker registry, set the following variables in the inventory file:
@@ -548,7 +552,7 @@ Using /etc/ansible/ansible.cfg as config file
 }
 Operations to perform:
   Synchronize unmigrated apps: solo, api, staticfiles, messages, channels, django_extensions, ui, rest_framework, polymorphic
-  Apply all migrations: sso, taggit, sessions, djcelery, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
+  Apply all migrations: sso, taggit, sessions, sites, kombu_transport_django, social_auth, contenttypes, auth, conf, main
 Synchronizing apps without migrations:
   Creating tables...
     Running deferred SQL...

Makefile

@@ -59,7 +59,7 @@ UI_RELEASE_FLAG_FILE = awx/ui/.release_built
 I18N_FLAG_FILE = .i18n_built
 
 .PHONY: awx-link clean clean-tmp clean-venv requirements requirements_dev \
-	develop refresh adduser migrate dbchange dbshell runserver celeryd \
+	develop refresh adduser migrate dbchange dbshell runserver \
 	receiver test test_unit test_ansible test_coverage coverage_html \
 	dev_build release_build release_clean sdist \
 	ui-docker-machine ui-docker ui-release ui-devel \
@@ -165,7 +165,7 @@ requirements_awx: virtualenv_awx
 	else \
 	    cat requirements/requirements.txt requirements/requirements_git.txt | $(VENV_BASE)/awx/bin/pip install $(PIP_OPTIONS) --no-binary $(SRC_ONLY_PKGS) --ignore-installed -r /dev/stdin ; \
 	fi
-	$(VENV_BASE)/awx/bin/pip uninstall --yes -r requirements/requirements_tower_uninstall.txt
+	#$(VENV_BASE)/awx/bin/pip uninstall --yes -r requirements/requirements_tower_uninstall.txt
 
 requirements_awx_dev:
 	$(VENV_BASE)/awx/bin/pip install -r requirements/requirements_dev.txt
@@ -233,7 +233,7 @@ server_noattach:
 	tmux new-session -d -s awx 'exec make uwsgi'
 	tmux rename-window 'AWX'
 	tmux select-window -t awx:0
-	tmux split-window -v 'exec make celeryd'
+	tmux split-window -v 'exec make dispatcher'
 	tmux new-window 'exec make daphne'
 	tmux select-window -t awx:1
 	tmux rename-window 'WebSockets'
@@ -265,12 +265,6 @@ honcho:
 	fi; \
 	honcho start -f tools/docker-compose/Procfile
 
-flower:
-	@if [ "$(VENV_BASE)" ]; then \
-		. $(VENV_BASE)/awx/bin/activate; \
-	fi; \
-	celery flower --address=0.0.0.0 --port=5555 --broker=amqp://guest:guest@$(RABBITMQ_HOST):5672//
-
 collectstatic:
 	@if [ "$(VENV_BASE)" ]; then \
 		. $(VENV_BASE)/awx/bin/activate; \
@@ -281,7 +275,7 @@ uwsgi: collectstatic
 	@if [ "$(VENV_BASE)" ]; then \
 		. $(VENV_BASE)/awx/bin/activate; \
 	fi; \
-    uwsgi -b 32768 --socket 127.0.0.1:8050 --module=awx.wsgi:application --home=/venv/awx --chdir=/awx_devel/ --vacuum --processes=5 --harakiri=120 --master --no-orphans --py-autoreload 1 --max-requests=1000 --stats /tmp/stats.socket --lazy-apps --logformat "%(addr) %(method) %(uri) - %(proto) %(status)" --hook-accepting1-once="exec:/bin/sh -c '[ -f /tmp/celery_pid ] && kill -1 `cat /tmp/celery_pid` || true'"
+    uwsgi -b 32768 --socket 127.0.0.1:8050 --module=awx.wsgi:application --home=/venv/awx --chdir=/awx_devel/ --vacuum --processes=5 --harakiri=120 --master --no-orphans --py-autoreload 1 --max-requests=1000 --stats /tmp/stats.socket --lazy-apps --logformat "%(addr) %(method) %(uri) - %(proto) %(status)" --hook-accepting1-once="exec:awx-manage run_dispatcher --reload"
 
 daphne:
 	@if [ "$(VENV_BASE)" ]; then \
@@ -302,13 +296,13 @@ runserver:
 	fi; \
 	$(PYTHON) manage.py runserver
 
-# Run to start the background celery worker for development.
-celeryd:
-	rm -f /tmp/celery_pid
+# Run to start the background task dispatcher for development.
+dispatcher:
 	@if [ "$(VENV_BASE)" ]; then \
 		. $(VENV_BASE)/awx/bin/activate; \
 	fi; \
-	celery worker -A awx -l DEBUG -B -Ofair --autoscale=100,4 --schedule=$(CELERY_SCHEDULE_FILE) -n celery@$(COMPOSE_HOST) --pidfile /tmp/celery_pid
+	$(PYTHON) manage.py run_dispatcher
+
 
 # Run to start the zeromq callback receiver
 receiver:
@@ -355,7 +349,7 @@ check: flake8 pep8 # pyflakes pylint
 awx-link:
 	cp -R /tmp/awx.egg-info /awx_devel/ || true
 	sed -i "s/placeholder/$(shell git describe --long | sed 's/\./\\./g')/" /awx_devel/awx.egg-info/PKG-INFO
-	cp /tmp/awx.egg-link /venv/awx/lib/python2.7/site-packages/awx.egg-link
+	cp -f /tmp/awx.egg-link /venv/awx/lib/python2.7/site-packages/awx.egg-link
 
 TEST_DIRS ?= awx/main/tests/unit awx/main/tests/functional awx/conf/tests awx/sso/tests
 
@@ -364,7 +358,7 @@ test:
 	@if [ "$(VENV_BASE)" ]; then \
 		. $(VENV_BASE)/awx/bin/activate; \
 	fi; \
-	py.test -n auto $(TEST_DIRS)
+	PYTHONDONTWRITEBYTECODE=1 py.test -p no:cacheprovider -n auto $(TEST_DIRS)
 	awx-manage check_migrations --dry-run --check  -n 'vNNN_missing_migration_file'
 
 test_combined: test_ansible test
@@ -466,7 +460,7 @@ $(I18N_FLAG_FILE): $(UI_DEPS_FLAG_FILE)
 ui-deps: $(UI_DEPS_FLAG_FILE)
 
 $(UI_DEPS_FLAG_FILE):
-	$(NPM_BIN) --unsafe-perm --prefix awx/ui install awx/ui
+	$(NPM_BIN) --unsafe-perm --prefix awx/ui install --no-save awx/ui
 	touch $(UI_DEPS_FLAG_FILE)
 
 ui-docker-machine: $(UI_DEPS_FLAG_FILE)
@@ -552,17 +546,27 @@ docker-isolated:
 	else \
 		docker exec "tools_isolated_1" bash -c "mkdir -p /root/.ssh && rm -f /root/.ssh/authorized_keys && echo $$(docker exec -t tools_awx_1 cat /root/.ssh/id_rsa.pub) >> /root/.ssh/authorized_keys"; \
 	fi
-	TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml -f tools/docker-isolated-override.yml up
+	CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml -f tools/docker-isolated-override.yml up
 
 # Docker Compose Development environment
 docker-compose: docker-auth
-	TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml up --no-recreate awx
+	CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml up --no-recreate awx
 
 docker-compose-cluster: docker-auth
-	TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose-cluster.yml up
+	CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose-cluster.yml up
 
 docker-compose-test: docker-auth
-	cd tools && TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose run --rm --service-ports awx /bin/bash
+	cd tools && CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose run --rm --service-ports awx /bin/bash
+
+docker-compose-runtest:
+	cd tools && CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose run --rm --service-ports awx /start_tests.sh
+
+docker-compose-build-swagger:
+	cd tools && CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose run --rm --service-ports awx /start_tests.sh swagger
+
+docker-compose-clean:
+	cd tools && CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose run --rm -w /awx_devel --service-ports awx make clean
+	cd tools && TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose rm -sf
 
 docker-compose-build: awx-devel-build
 
@@ -588,7 +592,7 @@ docker-refresh: docker-clean docker-compose
 
 # Docker Development Environment with Elastic Stack Connected
 docker-compose-elk: docker-auth
-	TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml -f tools/elastic/docker-compose.logstash-link.yml -f tools/elastic/docker-compose.elastic-override.yml up --no-recreate
+	CURRENT_UID=$(shell id -u) TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose.yml -f tools/elastic/docker-compose.logstash-link.yml -f tools/elastic/docker-compose.elastic-override.yml up --no-recreate
 
 docker-compose-cluster-elk: docker-auth
 	TAG=$(COMPOSE_TAG) DEV_DOCKER_TAG_BASE=$(DEV_DOCKER_TAG_BASE) docker-compose -f tools/docker-compose-cluster.yml -f tools/elastic/docker-compose.logstash-link-cluster.yml -f tools/elastic/docker-compose.elastic-override.yml up --no-recreate

README.md

@@ -1,7 +1,6 @@
-[![Run Status](https://api.shippable.com/projects/591c82a22f895107009e8b35/badge?branch=devel)](https://app.shippable.com/github/ansible/awx)
+[![Gated by Zuul](https://zuul-ci.org/gated.svg)](https://ansible.softwarefactory-project.io/zuul/status)
 
-AWX
-===
+<img src="https://raw.githubusercontent.com/ansible/awx-logos/master/awx/ui/client/assets/logo-login.svg?sanitize=true" width=200 alt="AWX" />
 
 AWX provides a web-based user interface, REST API, and task engine built on top of [Ansible](https://github.com/ansible/ansible). It is the upstream project for [Tower](https://www.ansible.com/tower), a commercial derivative of AWX.  
 
@@ -11,6 +10,8 @@ To learn more about using AWX, and Tower, view the [Tower docs site](http://docs
 
 The AWX Project Frequently Asked Questions can be found [here](https://www.ansible.com/awx-project-faq).
 
+The AWX logos and branding assets are covered by [our trademark guidelines](https://github.com/ansible/awx-logos/blob/master/TRADEMARKS.md).
+
 Contributing
 ------------
 

VERSION

@@ -1 +1 @@
-2.0.0
+2.1.0

awx/__init__.py

@@ -12,14 +12,6 @@ __version__ = get_distribution('awx').version
 __all__ = ['__version__']
 
 
-# Isolated nodes do not have celery installed
-try:
-    from .celery import app as celery_app # noqa
-    __all__.append('celery_app')
-except ImportError:
-    pass
-
-
 # Check for the presence/absence of "devonly" module to determine if running
 # from a source code checkout or release packaage.
 try:

awx/api/metadata.py

@@ -63,12 +63,15 @@ class Metadata(metadata.SimpleMetadata):
                 verbose_name = smart_text(opts.verbose_name)
                 field_info['help_text'] = field_help_text[field.field_name].format(verbose_name)
 
-            for model_field in serializer.Meta.model._meta.fields:
-                if field.field_name == model_field.name:
-                    field_info['filterable'] = True
-                    break
+            if field.field_name == 'type':
+                field_info['filterable'] = True
             else:
-                field_info['filterable'] = False
+                for model_field in serializer.Meta.model._meta.fields:
+                    if field.field_name == model_field.name:
+                        field_info['filterable'] = True
+                        break
+                else:
+                    field_info['filterable'] = False
 
         # Indicate if a field has a default value.
         # FIXME: Still isn't showing all default values?

awx/api/serializers.py

@@ -3011,7 +3011,7 @@ class JobTemplateSerializer(JobTemplateMixin, UnifiedJobTemplateSerializer, JobO
         fields = ('*', 'host_config_key', 'ask_diff_mode_on_launch', 'ask_variables_on_launch', 'ask_limit_on_launch', 'ask_tags_on_launch',
                   'ask_skip_tags_on_launch', 'ask_job_type_on_launch', 'ask_verbosity_on_launch', 'ask_inventory_on_launch',
                   'ask_credential_on_launch', 'survey_enabled', 'become_enabled', 'diff_mode',
-                  'allow_simultaneous', 'custom_virtualenv')
+                  'allow_simultaneous', 'custom_virtualenv', 'job_slice_count')
 
     def get_related(self, obj):
         res = super(JobTemplateSerializer, self).get_related(obj)
@@ -3028,6 +3028,7 @@ class JobTemplateSerializer(JobTemplateMixin, UnifiedJobTemplateSerializer, JobO
             labels = self.reverse('api:job_template_label_list', kwargs={'pk': obj.pk}),
             object_roles = self.reverse('api:job_template_object_roles_list', kwargs={'pk': obj.pk}),
             instance_groups = self.reverse('api:job_template_instance_groups_list', kwargs={'pk': obj.pk}),
+            slice_workflow_jobs = self.reverse('api:job_template_slice_workflow_jobs_list', kwargs={'pk': obj.pk}),
         ))
         if self.version > 1:
             res['copy'] = self.reverse('api:job_template_copy', kwargs={'pk': obj.pk})
@@ -3123,7 +3124,7 @@ class JobSerializer(UnifiedJobSerializer, JobOptionsSerializer):
                   'ask_variables_on_launch', 'ask_limit_on_launch', 'ask_tags_on_launch', 'ask_skip_tags_on_launch',
                   'ask_job_type_on_launch', 'ask_verbosity_on_launch', 'ask_inventory_on_launch',
                   'ask_credential_on_launch', 'allow_simultaneous', 'artifacts', 'scm_revision',
-                  'instance_group', 'diff_mode')
+                  'instance_group', 'diff_mode', 'job_slice_number', 'job_slice_count')
 
     def get_related(self, obj):
         res = super(JobSerializer, self).get_related(obj)
@@ -3590,6 +3591,7 @@ class WorkflowJobSerializer(LabelsListMixin, UnifiedJobSerializer):
     class Meta:
         model = WorkflowJob
         fields = ('*', 'workflow_job_template', 'extra_vars', 'allow_simultaneous',
+                  'job_template', 'is_sliced_job',
                   '-execution_node', '-event_processing_finished', '-controller_node',)
 
     def get_related(self, obj):
@@ -3598,6 +3600,8 @@ class WorkflowJobSerializer(LabelsListMixin, UnifiedJobSerializer):
             res['workflow_job_template'] = self.reverse('api:workflow_job_template_detail',
                                                         kwargs={'pk': obj.workflow_job_template.pk})
             res['notifications'] = self.reverse('api:workflow_job_notifications_list', kwargs={'pk': obj.pk})
+        if obj.job_template_id:
+            res['job_template'] = self.reverse('api:job_template_detail', kwargs={'pk': obj.job_template_id})
         res['workflow_nodes'] = self.reverse('api:workflow_job_workflow_nodes_list', kwargs={'pk': obj.pk})
         res['labels'] = self.reverse('api:workflow_job_label_list', kwargs={'pk': obj.pk})
         res['activity_stream'] = self.reverse('api:workflow_job_activity_stream_list', kwargs={'pk': obj.pk})
@@ -4535,13 +4539,13 @@ class SchedulePreviewSerializer(BaseSerializer):
     # - COUNT > 999
     def validate_rrule(self, value):
         rrule_value = value
-        multi_by_month_day = ".*?BYMONTHDAY[\:\=][0-9]+,-*[0-9]+"
-        multi_by_month = ".*?BYMONTH[\:\=][0-9]+,[0-9]+"
-        by_day_with_numeric_prefix = ".*?BYDAY[\:\=][0-9]+[a-zA-Z]{2}"
-        match_count = re.match(".*?(COUNT\=[0-9]+)", rrule_value)
-        match_multiple_dtstart = re.findall(".*?(DTSTART(;[^:]+)?\:[0-9]+T[0-9]+Z?)", rrule_value)
-        match_native_dtstart = re.findall(".*?(DTSTART:[0-9]+T[0-9]+) ", rrule_value)
-        match_multiple_rrule = re.findall(".*?(RRULE\:)", rrule_value)
+        multi_by_month_day = r".*?BYMONTHDAY[\:\=][0-9]+,-*[0-9]+"
+        multi_by_month = r".*?BYMONTH[\:\=][0-9]+,[0-9]+"
+        by_day_with_numeric_prefix = r".*?BYDAY[\:\=][0-9]+[a-zA-Z]{2}"
+        match_count = re.match(r".*?(COUNT\=[0-9]+)", rrule_value)
+        match_multiple_dtstart = re.findall(r".*?(DTSTART(;[^:]+)?\:[0-9]+T[0-9]+Z?)", rrule_value)
+        match_native_dtstart = re.findall(r".*?(DTSTART:[0-9]+T[0-9]+) ", rrule_value)
+        match_multiple_rrule = re.findall(r".*?(RRULE\:)", rrule_value)
         if not len(match_multiple_dtstart):
             raise serializers.ValidationError(_('Valid DTSTART required in rrule. Value should start with: DTSTART:YYYYMMDDTHHMMSSZ'))
         if len(match_native_dtstart):

awx/api/templates/api/inventory_script_view.md

@@ -26,6 +26,9 @@ string of `?all=1` to return all hosts, including disabled ones.
 Specify a query string of `?towervars=1` to add variables
 to the hostvars of each host that specifies its enabled state and database ID.
 
+Specify a query string of `?subset=slice2of5` to produce an inventory that
+has a restricted number of hosts according to the rules of job slicing.
+
 To apply multiple query strings, join them with the `&` character, like `?hostvars=1&all=1`.
 
 ## Host Response

awx/api/urls/job_template.py

@@ -8,6 +8,7 @@ from awx.api.views import (
     JobTemplateDetail,
     JobTemplateLaunch,
     JobTemplateJobsList,
+    JobTemplateSliceWorkflowJobsList,
     JobTemplateCallback,
     JobTemplateSchedulesList,
     JobTemplateSurveySpec,
@@ -28,6 +29,7 @@ urls = [
     url(r'^(?P<pk>[0-9]+)/$', JobTemplateDetail.as_view(), name='job_template_detail'),
     url(r'^(?P<pk>[0-9]+)/launch/$', JobTemplateLaunch.as_view(), name='job_template_launch'),
     url(r'^(?P<pk>[0-9]+)/jobs/$', JobTemplateJobsList.as_view(), name='job_template_jobs_list'),
+    url(r'^(?P<pk>[0-9]+)/slice_workflow_jobs/$', JobTemplateSliceWorkflowJobsList.as_view(), name='job_template_slice_workflow_jobs_list'),
     url(r'^(?P<pk>[0-9]+)/callback/$', JobTemplateCallback.as_view(), name='job_template_callback'),
     url(r'^(?P<pk>[0-9]+)/schedules/$', JobTemplateSchedulesList.as_view(), name='job_template_schedules_list'),
     url(r'^(?P<pk>[0-9]+)/survey_spec/$', JobTemplateSurveySpec.as_view(), name='job_template_survey_spec'),

awx/api/views/__init__.py

@@ -19,7 +19,7 @@ import six
 from django.conf import settings
 from django.core.exceptions import FieldError, ObjectDoesNotExist
 from django.db.models import Q, Count
-from django.db import IntegrityError, transaction
+from django.db import IntegrityError, transaction, connection
 from django.shortcuts import get_object_or_404
 from django.utils.encoding import smart_text
 from django.utils.safestring import mark_safe
@@ -599,7 +599,7 @@ class InstanceGroupInstanceList(InstanceGroupMembershipMixin, SubListAttachDetac
     search_fields = ('hostname',)
 
 
-class ScheduleList(ListAPIView):
+class ScheduleList(ListCreateAPIView):
 
     view_name = _("Schedules")
     model = Schedule
@@ -1291,7 +1291,7 @@ class ProjectUpdateNotificationsList(SubListAPIView):
     search_fields = ('subject', 'notification_type', 'body',)
 
 
-class ProjectUpdateScmInventoryUpdates(SubListCreateAPIView):
+class ProjectUpdateScmInventoryUpdates(SubListAPIView):
 
     view_name = _("Project Update SCM Inventory Updates")
     model = InventoryUpdate
@@ -1484,10 +1484,11 @@ class OAuth2TokenActivityStreamList(ActivityStreamEnforcementMixin, SubListAPIVi
     search_fields = ('changes',)
 
 
-class UserTeamsList(ListAPIView):
+class UserTeamsList(SubListAPIView):
 
-    model = User
+    model = Team
     serializer_class = TeamSerializer
+    parent_model = User
 
     def get_queryset(self):
         u = get_object_or_404(User, pk=self.kwargs['pk'])
@@ -1665,7 +1666,7 @@ class CredentialTypeDetail(RetrieveUpdateDestroyAPIView):
         return super(CredentialTypeDetail, self).destroy(request, *args, **kwargs)
 
 
-class CredentialTypeCredentialList(SubListAPIView):
+class CredentialTypeCredentialList(SubListCreateAPIView):
 
     model = Credential
     parent_model = CredentialType
@@ -2451,6 +2452,16 @@ class InventoryScriptView(RetrieveAPIView):
         hostvars = bool(request.query_params.get('hostvars', ''))
         towervars = bool(request.query_params.get('towervars', ''))
         show_all = bool(request.query_params.get('all', ''))
+        subset = request.query_params.get('subset', '')
+        if subset:
+            if not isinstance(subset, six.string_types):
+                raise ParseError(_('Inventory subset argument must be a string.'))
+            if subset.startswith('slice'):
+                slice_number, slice_count = Inventory.parse_slice_params(subset)
+            else:
+                raise ParseError(_('Subset does not use any supported syntax.'))
+        else:
+            slice_number, slice_count = 1, 1
         if hostname:
             hosts_q = dict(name=hostname)
             if not show_all:
@@ -2460,7 +2471,8 @@ class InventoryScriptView(RetrieveAPIView):
         return Response(obj.get_script_data(
             hostvars=hostvars,
             towervars=towervars,
-            show_all=show_all
+            show_all=show_all,
+            slice_number=slice_number, slice_count=slice_count
         ))
 
 
@@ -2911,9 +2923,14 @@ class JobTemplateLaunch(RetrieveAPIView):
             return Response(data, status=status.HTTP_400_BAD_REQUEST)
         else:
             data = OrderedDict()
-            data['job'] = new_job.id
-            data['ignored_fields'] = self.sanitize_for_response(ignored_fields)
-            data.update(JobSerializer(new_job, context=self.get_serializer_context()).to_representation(new_job))
+            if isinstance(new_job, WorkflowJob):
+                data['workflow_job'] = new_job.id
+                data['ignored_fields'] = self.sanitize_for_response(ignored_fields)
+                data.update(WorkflowJobSerializer(new_job, context=self.get_serializer_context()).to_representation(new_job))
+            else:
+                data['job'] = new_job.id
+                data['ignored_fields'] = self.sanitize_for_response(ignored_fields)
+                data.update(JobSerializer(new_job, context=self.get_serializer_context()).to_representation(new_job))
             headers = {'Location': new_job.get_absolute_url(request)}
             return Response(data, status=status.HTTP_201_CREATED, headers=headers)
 
@@ -2959,6 +2976,16 @@ class JobTemplateSurveySpec(GenericAPIView):
     obj_permission_type = 'admin'
     serializer_class = EmptySerializer
 
+    ALLOWED_TYPES = {
+        'text': six.string_types,
+        'textarea': six.string_types,
+        'password': six.string_types,
+        'multiplechoice': six.string_types,
+        'multiselect': six.string_types,
+        'integer': int,
+        'float': float
+    }
+
     def get(self, request, *args, **kwargs):
         obj = self.get_object()
         if not feature_enabled('surveys'):
@@ -2985,7 +3012,8 @@ class JobTemplateSurveySpec(GenericAPIView):
         obj.save(update_fields=['survey_spec'])
         return Response()
 
-    def _validate_spec_data(self, new_spec, old_spec):
+    @staticmethod
+    def _validate_spec_data(new_spec, old_spec):
         schema_errors = {}
         for field, expect_type, type_label in [
                 ('name', six.string_types, 'string'),
@@ -3006,40 +3034,75 @@ class JobTemplateSurveySpec(GenericAPIView):
         variable_set = set()
         old_spec_dict = JobTemplate.pivot_spec(old_spec)
         for idx, survey_item in enumerate(new_spec["spec"]):
+            context = dict(
+                idx=six.text_type(idx),
+                survey_item=survey_item
+            )
+            # General element validation
             if not isinstance(survey_item, dict):
                 return Response(dict(error=_("Survey question %s is not a json object.") % str(idx)), status=status.HTTP_400_BAD_REQUEST)
-            if "type" not in survey_item:
-                return Response(dict(error=_("'type' missing from survey question %s.") % str(idx)), status=status.HTTP_400_BAD_REQUEST)
-            if "question_name" not in survey_item:
-                return Response(dict(error=_("'question_name' missing from survey question %s.") % str(idx)), status=status.HTTP_400_BAD_REQUEST)
-            if "variable" not in survey_item:
-                return Response(dict(error=_("'variable' missing from survey question %s.") % str(idx)), status=status.HTTP_400_BAD_REQUEST)
+            for field_name in ['type', 'question_name', 'variable', 'required']:
+                if field_name not in survey_item:
+                    return Response(dict(error=_("'{field_name}' missing from survey question {idx}").format(
+                        field_name=field_name, **context
+                    )), status=status.HTTP_400_BAD_REQUEST)
+                val = survey_item[field_name]
+                allow_types = six.string_types
+                type_label = 'string'
+                if field_name == 'required':
+                    allow_types = bool
+                    type_label = 'boolean'
+                if not isinstance(val, allow_types):
+                    return Response(dict(error=_("'{field_name}' in survey question {idx} expected to be {type_label}.").format(
+                        field_name=field_name, type_label=type_label, **context
+                    )))
             if survey_item['variable'] in variable_set:
                 return Response(dict(error=_("'variable' '%(item)s' duplicated in survey question %(survey)s.") % {
                     'item': survey_item['variable'], 'survey': str(idx)}), status=status.HTTP_400_BAD_REQUEST)
             else:
                 variable_set.add(survey_item['variable'])
-            if "required" not in survey_item:
-                return Response(dict(error=_("'required' missing from survey question %s.") % str(idx)), status=status.HTTP_400_BAD_REQUEST)
 
-            if survey_item["type"] == "password" and "default" in survey_item:
-                if not isinstance(survey_item['default'], six.string_types):
+            # Type-specific validation
+            # validate question type <-> default type
+            qtype = survey_item["type"]
+            if qtype not in JobTemplateSurveySpec.ALLOWED_TYPES:
+                return Response(dict(error=_(
+                    "'{survey_item[type]}' in survey question {idx} is not one of '{allowed_types}' allowed question types."
+                ).format(
+                    allowed_types=', '.join(JobTemplateSurveySpec.ALLOWED_TYPES.keys()), **context
+                )))
+            if 'default' in survey_item:
+                if not isinstance(survey_item['default'], JobTemplateSurveySpec.ALLOWED_TYPES[qtype]):
+                    type_label = 'string'
+                    if qtype in ['integer', 'float']:
+                        type_label = qtype
                     return Response(dict(error=_(
-                        "Value {question_default} for '{variable_name}' expected to be a string."
+                        "Default value {survey_item[default]} in survey question {idx} expected to be {type_label}."
                     ).format(
-                        question_default=survey_item["default"], variable_name=survey_item["variable"])
-                    ), status=status.HTTP_400_BAD_REQUEST)
+                        type_label=type_label, **context
+                    )), status=status.HTTP_400_BAD_REQUEST)
+            # additional type-specific properties, the UI provides these even
+            # if not applicable to the question, TODO: request that they not do this
+            for key in ['min', 'max']:
+                if key in survey_item:
+                    if survey_item[key] is not None and (not isinstance(survey_item[key], int)):
+                        return Response(dict(error=_(
+                            "The {min_or_max} limit in survey question {idx} expected to be integer."
+                        ).format(min_or_max=key, **context)))
+            if qtype in ['multiplechoice', 'multiselect'] and 'choices' not in survey_item:
+                return Response(dict(error=_(
+                    "Survey question {idx} of type {survey_item[type]} must specify choices.".format(**context)
+                )))
 
+            # Process encryption substitution
             if ("default" in survey_item and isinstance(survey_item['default'], six.string_types) and
                     survey_item['default'].startswith('$encrypted$')):
                 # Submission expects the existence of encrypted DB value to replace given default
-                if survey_item["type"] != "password":
+                if qtype != "password":
                     return Response(dict(error=_(
                         "$encrypted$ is a reserved keyword for password question defaults, "
-                        "survey question {question_position} is type {question_type}."
-                    ).format(
-                        question_position=str(idx), question_type=survey_item["type"])
-                    ), status=status.HTTP_400_BAD_REQUEST)
+                        "survey question {idx} is type {survey_item[type]}."
+                    ).format(**context)), status=status.HTTP_400_BAD_REQUEST)
                 old_element = old_spec_dict.get(survey_item['variable'], {})
                 encryptedish_default_exists = False
                 if 'default' in old_element:
@@ -3051,10 +3114,10 @@ class JobTemplateSurveySpec(GenericAPIView):
                             encryptedish_default_exists = True
                 if not encryptedish_default_exists:
                     return Response(dict(error=_(
-                        "$encrypted$ is a reserved keyword, may not be used for new default in position {question_position}."
-                    ).format(question_position=str(idx))), status=status.HTTP_400_BAD_REQUEST)
+                        "$encrypted$ is a reserved keyword, may not be used for new default in position {idx}."
+                    ).format(**context)), status=status.HTTP_400_BAD_REQUEST)
                 survey_item['default'] = old_element['default']
-            elif survey_item["type"] == "password" and 'default' in survey_item:
+            elif qtype == "password" and 'default' in survey_item:
                 # Submission provides new encrypted default
                 survey_item['default'] = encrypt_value(survey_item['default'])
 
@@ -3125,8 +3188,8 @@ class JobTemplateCredentialsList(SubListCreateAttachDetachAPIView):
 
     def is_valid_relation(self, parent, sub, created=False):
         if sub.unique_hash() in [cred.unique_hash() for cred in parent.credentials.all()]:
-            return {"error": _("Cannot assign multiple {credential_type} credentials.".format(
-                credential_type=sub.unique_hash(display=True)))}
+            return {"error": _("Cannot assign multiple {credential_type} credentials.").format(
+                credential_type=sub.unique_hash(display=True))}
         kind = sub.credential_type.kind
         if kind not in ('ssh', 'vault', 'cloud', 'net'):
             return {'error': _('Cannot assign a Credential of kind `{}`.').format(kind)}
@@ -3315,10 +3378,11 @@ class JobTemplateCallback(GenericAPIView):
         if extra_vars is not None and job_template.ask_variables_on_launch:
             extra_vars_redacted, removed = extract_ansible_vars(extra_vars)
             kv['extra_vars'] = extra_vars_redacted
+        kv['_prevent_slicing'] = True  # will only run against 1 host, so no point
         with transaction.atomic():
             job = job_template.create_job(**kv)
 
-        # Send a signal to celery that the job should be started.
+        # Send a signal to signify that the job should be started.
         result = job.signal_start(inventory_sources_already_updated=inventory_sources_already_updated)
         if not result:
             data = dict(msg=_('Error starting job!'))
@@ -3346,6 +3410,15 @@ class JobTemplateJobsList(SubListCreateAPIView):
         return methods
 
 
+class JobTemplateSliceWorkflowJobsList(SubListCreateAPIView):
+
+    model = WorkflowJob
+    serializer_class = WorkflowJobListSerializer
+    parent_model = JobTemplate
+    relationship = 'slice_workflow_jobs'
+    parent_key = 'job_template'
+
+
 class JobTemplateInstanceGroupsList(SubListAttachDetachAPIView):
 
     model = InstanceGroup
@@ -3438,10 +3511,6 @@ class WorkflowJobTemplateNodeChildrenBaseList(WorkflowsEnforcementMixin, Enforce
         return getattr(parent, self.relationship).all()
 
     def is_valid_relation(self, parent, sub, created=False):
-        mutex_list = ('success_nodes', 'failure_nodes') if self.relationship == 'always_nodes' else ('always_nodes',)
-        for relation in mutex_list:
-            if getattr(parent, relation).all().exists():
-                return {'Error': _('Cannot associate {0} when {1} have been associated.').format(self.relationship, relation)}
 
         if created:
             return None
@@ -3642,6 +3711,8 @@ class WorkflowJobRelaunch(WorkflowsEnforcementMixin, GenericAPIView):
 
     def post(self, request, *args, **kwargs):
         obj = self.get_object()
+        if obj.is_sliced_job and not obj.job_template_id:
+            raise ParseError(_('Cannot relaunch slice workflow job orphaned from job template.'))
         new_workflow_job = obj.create_relaunch_workflow_job()
         new_workflow_job.signal_start()
 

awx/api/views/mixin.py

@@ -101,7 +101,9 @@ class UnifiedJobDeletionMixin(object):
 
 class InstanceGroupMembershipMixin(object):
     '''
-    Manages signaling celery to reload its queue configuration on Instance Group membership changes
+    This mixin overloads attach/detach so that it calls InstanceGroup.save(),
+    triggering a background recalculation of policy-based instance group
+    membership.
     '''
     def attach(self, request, *args, **kwargs):
         response = super(InstanceGroupMembershipMixin, self).attach(request, *args, **kwargs)

awx/celery.py

@@ -1,25 +0,0 @@
-
-# Copyright (c) 2017 Ansible, Inc.
-# All Rights Reserved.
-
-from __future__ import absolute_import, unicode_literals
-
-import os
-from celery import Celery
-from django.conf import settings # noqa
-
-
-try:
-    import awx.devonly # noqa
-    MODE = 'development'
-except ImportError: # pragma: no cover
-    MODE = 'production'
-
-os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'awx.settings.%s' % MODE)
-
-app = Celery('awx')
-app.config_from_object('django.conf:settings')
-app.autodiscover_tasks(lambda: settings.INSTALLED_APPS)
-
-if __name__ == '__main__':
-    app.start()

awx/conf/settings.py

@@ -2,11 +2,13 @@
 from collections import namedtuple
 import contextlib
 import logging
+import re
 import sys
 import threading
 import time
 import StringIO
 import traceback
+import urllib
 
 import six
 
@@ -60,6 +62,15 @@ SETTING_CACHE_DEFAULTS = True
 __all__ = ['SettingsWrapper', 'get_settings_to_cache', 'SETTING_CACHE_NOTSET']
 
 
+def normalize_broker_url(value):
+    parts = value.rsplit('@', 1)
+    match = re.search('(amqp://[^:]+:)(.*)', parts[0])
+    if match:
+        prefix, password = match.group(1), match.group(2)
+        parts[0] = prefix + urllib.quote(password)
+    return '@'.join(parts)
+
+
 @contextlib.contextmanager
 def _ctit_db_wrapper(trans_safe=False):
     '''
@@ -115,7 +126,8 @@ def _ctit_db_wrapper(trans_safe=False):
             bottom_stack.close()
             # Log the combined stack
             if trans_safe:
-                logger.warning('Database settings are not available, using defaults, error:\n{}'.format(tb_string))
+                if 'check_migrations' not in sys.argv:
+                    logger.warning('Database settings are not available, using defaults, error:\n{}'.format(tb_string))
             else:
                 logger.error('Error modifying something related to database settings.\n{}'.format(tb_string))
     finally:
@@ -444,7 +456,16 @@ class SettingsWrapper(UserSettingsHolder):
                 value = self._get_local(name)
         if value is not empty:
             return value
-        return self._get_default(name)
+        value = self._get_default(name)
+        # sometimes users specify RabbitMQ passwords that contain
+        # unescaped : and @ characters that confused urlparse, e.g.,
+        # amqp://guest:a@ns:ibl3#@localhost:5672//
+        #
+        # detect these scenarios, and automatically escape the user's
+        # password so it just works
+        if name == 'BROKER_URL':
+            value = normalize_broker_url(value)
+        return value
 
     def _set_local(self, name, value):
         field = self.registry.get_setting_field(name)

awx/main/access.py

@@ -1789,7 +1789,7 @@ class WorkflowJobNodeAccess(BaseAccess):
 
     def filtered_queryset(self):
         return self.model.objects.filter(
-            workflow_job__workflow_job_template__in=WorkflowJobTemplate.accessible_objects(
+            workflow_job__unified_job_template__in=UnifiedJobTemplate.accessible_pk_qs(
                 self.user, 'read_role'))
 
     @check_superuser
@@ -1915,7 +1915,7 @@ class WorkflowJobAccess(BaseAccess):
 
     def filtered_queryset(self):
         return WorkflowJob.objects.filter(
-            workflow_job_template__in=WorkflowJobTemplate.accessible_objects(
+            unified_job_template__in=UnifiedJobTemplate.accessible_pk_qs(
                 self.user, 'read_role'))
 
     def can_add(self, data):
@@ -1947,9 +1947,11 @@ class WorkflowJobAccess(BaseAccess):
         if self.user.is_superuser:
             return True
 
-        wfjt = obj.workflow_job_template
+        template = obj.workflow_job_template
+        if not template and obj.job_template_id:
+            template = obj.job_template
         # only superusers can relaunch orphans
-        if not wfjt:
+        if not template:
             return False
 
         # If job was launched by another user, it could have survey passwords
@@ -1967,7 +1969,7 @@ class WorkflowJobAccess(BaseAccess):
                 return False
 
         # execute permission to WFJT is mandatory for any relaunch
-        return (self.user in wfjt.execute_role)
+        return (self.user in template.execute_role)
 
     def can_recreate(self, obj):
         node_qs = obj.workflow_job_nodes.all().prefetch_related('inventory', 'credentials', 'unified_job_template')

awx/main/constants.py

@@ -29,3 +29,11 @@ STANDARD_INVENTORY_UPDATE_ENV = {
 CAN_CANCEL = ('new', 'pending', 'waiting', 'running')
 ACTIVE_STATES = CAN_CANCEL
 CENSOR_VALUE = '************'
+ENV_BLACKLIST = frozenset((
+    'VIRTUAL_ENV', 'PATH', 'PYTHONPATH', 'PROOT_TMP_DIR', 'JOB_ID',
+    'INVENTORY_ID', 'INVENTORY_SOURCE_ID', 'INVENTORY_UPDATE_ID',
+    'AD_HOC_COMMAND_ID', 'REST_API_URL', 'REST_API_TOKEN', 'MAX_EVENT_RES',
+    'CALLBACK_QUEUE', 'CALLBACK_CONNECTION', 'CACHE',
+    'JOB_CALLBACK_DEBUG', 'INVENTORY_HOSTVARS', 'FACT_QUEUE',
+    'AWX_HOST', 'PROJECT_REVISION'
+))

awx/main/dispatch/__init__.py

@@ -0,0 +1,5 @@
+from django.conf import settings
+
+
+def get_local_queuename():
+    return settings.CLUSTER_HOST_ID.encode('utf-8')

awx/main/dispatch/control.py

@@ -0,0 +1,58 @@
+import logging
+import socket
+
+from django.conf import settings
+
+from awx.main.dispatch import get_local_queuename
+from kombu import Connection, Queue, Exchange, Producer, Consumer
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+class Control(object):
+
+    services = ('dispatcher', 'callback_receiver')
+    result = None
+
+    def __init__(self, service, host=None):
+        if service not in self.services:
+            raise RuntimeError('{} must be in {}'.format(service, self.services))
+        self.service = service
+        self.queuename = host or get_local_queuename()
+        self.queue = Queue(self.queuename, Exchange(self.queuename), routing_key=self.queuename)
+
+    def publish(self, msg, conn, **kwargs):
+        producer = Producer(
+            exchange=self.queue.exchange,
+            channel=conn,
+            routing_key=self.queuename
+        )
+        producer.publish(msg, expiration=5, **kwargs)
+
+    def status(self, *args, **kwargs):
+        return self.control_with_reply('status', *args, **kwargs)
+
+    def running(self, *args, **kwargs):
+        return self.control_with_reply('running', *args, **kwargs)
+
+    def control_with_reply(self, command, timeout=5):
+        logger.warn('checking {} {} for {}'.format(self.service, command, self.queuename))
+        reply_queue = Queue(name="amq.rabbitmq.reply-to")
+        self.result = None
+        with Connection(settings.BROKER_URL) as conn:
+            with Consumer(conn, reply_queue, callbacks=[self.process_message], no_ack=True):
+                self.publish({'control': command}, conn, reply_to='amq.rabbitmq.reply-to')
+                try:
+                    conn.drain_events(timeout=timeout)
+                except socket.timeout:
+                    logger.error('{} did not reply within {}s'.format(self.service, timeout))
+                    raise
+        return self.result
+
+    def control(self, msg, **kwargs):
+        with Connection(settings.BROKER_URL) as conn:
+            self.publish(msg, conn)
+
+    def process_message(self, body, message):
+        self.result = body
+        message.ack()

awx/main/dispatch/pool.py

@@ -0,0 +1,405 @@
+import logging
+import os
+import sys
+import random
+import traceback
+from uuid import uuid4
+
+import collections
+from multiprocessing import Process
+from multiprocessing import Queue as MPQueue
+from Queue import Full as QueueFull, Empty as QueueEmpty
+
+from django.conf import settings
+from django.db import connection as django_connection, connections
+from django.core.cache import cache as django_cache
+from jinja2 import Template
+import psutil
+
+from awx.main.models import UnifiedJob
+from awx.main.dispatch import reaper
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+class PoolWorker(object):
+    '''
+    Used to track a worker child process and its pending and finished messages.
+
+    This class makes use of two distinct multiprocessing.Queues to track state:
+
+    - self.queue: this is a queue which represents pending messages that should
+                  be handled by this worker process; as new AMQP messages come
+                  in, a pool will put() them into this queue; the child
+                  process that is forked will get() from this queue and handle
+                  received messages in an endless loop
+    - self.finished: this is a queue which the worker process uses to signal
+                     that it has finished processing a message
+
+    When a message is put() onto this worker, it is tracked in
+    self.managed_tasks.
+
+    Periodically, the worker will call .calculate_managed_tasks(), which will
+    cause messages in self.finished to be removed from self.managed_tasks.
+
+    In this way, self.managed_tasks represents a view of the messages assigned
+    to a specific process.  The message at [0] is the least-recently inserted
+    message, and it represents what the worker is running _right now_
+    (self.current_task).
+
+    A worker is "busy" when it has at least one message in self.managed_tasks.
+    It is "idle" when self.managed_tasks is empty.
+    '''
+
+    def __init__(self, queue_size, target, args):
+        self.messages_sent = 0
+        self.messages_finished = 0
+        self.managed_tasks = collections.OrderedDict()
+        self.finished = MPQueue(queue_size)
+        self.queue = MPQueue(queue_size)
+        self.process = Process(target=target, args=(self.queue, self.finished) + args)
+        self.process.daemon = True
+
+    def start(self):
+        self.process.start()
+
+    def put(self, body):
+        uuid = '?'
+        if isinstance(body, dict):
+            if not body.get('uuid'):
+                body['uuid'] = str(uuid4())
+            uuid = body['uuid']
+        logger.debug('delivered {} to worker[{}] qsize {}'.format(
+            uuid, self.pid, self.qsize
+        ))
+        self.managed_tasks[uuid] = body
+        self.queue.put(body, block=True, timeout=5)
+        self.messages_sent += 1
+        self.calculate_managed_tasks()
+
+    def quit(self):
+        '''
+        Send a special control message to the worker that tells it to exit
+        gracefully.
+        '''
+        self.queue.put('QUIT')
+
+    @property
+    def pid(self):
+        return self.process.pid
+
+    @property
+    def qsize(self):
+        return self.queue.qsize()
+
+    @property
+    def alive(self):
+        return self.process.is_alive()
+
+    @property
+    def mb(self):
+        if self.alive:
+            return '{:0.3f}'.format(
+                psutil.Process(self.pid).memory_info().rss / 1024.0 / 1024.0
+            )
+        return '0'
+
+    @property
+    def exitcode(self):
+        return str(self.process.exitcode)
+
+    def calculate_managed_tasks(self):
+        # look to see if any tasks were finished
+        finished = []
+        for _ in range(self.finished.qsize()):
+            try:
+                finished.append(self.finished.get(block=False))
+            except QueueEmpty:
+                break  # qsize is not always _totally_ up to date
+
+        # if any tasks were finished, removed them from the managed tasks for
+        # this worker
+        for uuid in finished:
+            self.messages_finished += 1
+            del self.managed_tasks[uuid]
+
+    @property
+    def current_task(self):
+        self.calculate_managed_tasks()
+        # the task at [0] is the one that's running right now (or is about to
+        # be running)
+        if len(self.managed_tasks):
+            return self.managed_tasks[self.managed_tasks.keys()[0]]
+
+        return None
+
+    @property
+    def orphaned_tasks(self):
+        orphaned = []
+        if not self.alive:
+            # if this process had a running task that never finished,
+            # requeue its error callbacks
+            current_task = self.current_task
+            if isinstance(current_task, dict):
+                orphaned.extend(current_task.get('errbacks', []))
+
+            # if this process has any pending messages requeue them
+            for _ in range(self.qsize):
+                try:
+                    message = self.queue.get(block=False)
+                    if message != 'QUIT':
+                        orphaned.append(message)
+                except QueueEmpty:
+                    break  # qsize is not always _totally_ up to date
+            if len(orphaned):
+                logger.error(
+                    'requeuing {} messages from gone worker pid:{}'.format(
+                        len(orphaned), self.pid
+                    )
+                )
+        return orphaned
+
+    @property
+    def busy(self):
+        self.calculate_managed_tasks()
+        return len(self.managed_tasks) > 0
+
+    @property
+    def idle(self):
+        return not self.busy
+
+
+class WorkerPool(object):
+    '''
+    Creates a pool of forked PoolWorkers.
+
+    As WorkerPool.write(...) is called (generally, by a kombu consumer
+    implementation when it receives an AMQP message), messages are passed to
+    one of the multiprocessing Queues where some work can be done on them.
+
+    class MessagePrinter(awx.main.dispatch.worker.BaseWorker):
+
+        def perform_work(self, body):
+            print body
+
+    pool = WorkerPool(min_workers=4)  # spawn four worker processes
+    pool.init_workers(MessagePrint().work_loop)
+    pool.write(
+        0,  # preferred worker 0
+        'Hello, World!'
+    )
+    '''
+
+    debug_meta = ''
+
+    def __init__(self, min_workers=None, queue_size=None):
+        self.name = settings.CLUSTER_HOST_ID
+        self.pid = os.getpid()
+        self.min_workers = min_workers or settings.JOB_EVENT_WORKERS
+        self.queue_size = queue_size or settings.JOB_EVENT_MAX_QUEUE_SIZE
+        self.workers = []
+
+    def __len__(self):
+        return len(self.workers)
+
+    def init_workers(self, target, *target_args):
+        self.target = target
+        self.target_args = target_args
+        for idx in range(self.min_workers):
+            self.up()
+
+    def up(self):
+        idx = len(self.workers)
+        # It's important to close these because we're _about_ to fork, and we
+        # don't want the forked processes to inherit the open sockets
+        # for the DB and memcached connections (that way lies race conditions)
+        django_connection.close()
+        django_cache.close()
+        worker = PoolWorker(self.queue_size, self.target, (idx,) + self.target_args)
+        self.workers.append(worker)
+        try:
+            worker.start()
+        except Exception:
+            logger.exception('could not fork')
+        else:
+            logger.warn('scaling up worker pid:{}'.format(worker.pid))
+        return idx, worker
+
+    def debug(self, *args, **kwargs):
+        self.cleanup()
+        tmpl = Template(
+            '{{ pool.name }}[pid:{{ pool.pid }}] workers total={{ workers|length }} {{ meta }} \n'
+            '{% for w in workers %}'
+            '.  worker[pid:{{ w.pid }}]{% if not w.alive %} GONE exit={{ w.exitcode }}{% endif %}'
+            ' sent={{ w.messages_sent }}'
+            ' finished={{ w.messages_finished }}'
+            ' qsize={{ w.managed_tasks|length }}'
+            ' rss={{ w.mb }}MB'
+            '{% for task in w.managed_tasks.values() %}'
+            '\n     - {% if loop.index0 == 0 %}running {% else %}queued {% endif %}'
+            '{{ task["uuid"] }} '
+            '{% if "task" in task %}'
+            '{{ task["task"].rsplit(".", 1)[-1] }}'
+            # don't print kwargs, they often contain launch-time secrets
+            '(*{{ task.get("args", []) }})'
+            '{% endif %}'
+            '{% endfor %}'
+            '{% if not w.managed_tasks|length %}'
+            ' [IDLE]'
+            '{% endif %}'
+            '\n'
+            '{% endfor %}'
+        )
+        return tmpl.render(pool=self, workers=self.workers, meta=self.debug_meta)
+
+    def write(self, preferred_queue, body):
+        queue_order = sorted(range(len(self.workers)), cmp=lambda x, y: -1 if x==preferred_queue else 0)
+        write_attempt_order = []
+        for queue_actual in queue_order:
+            try:
+                self.workers[queue_actual].put(body)
+                return queue_actual
+            except QueueFull:
+                pass
+            except Exception:
+                tb = traceback.format_exc()
+                logger.warn("could not write to queue %s" % preferred_queue)
+                logger.warn("detail: {}".format(tb))
+            write_attempt_order.append(preferred_queue)
+        logger.warn("could not write payload to any queue, attempted order: {}".format(write_attempt_order))
+        return None
+
+    def stop(self, signum):
+        try:
+            for worker in self.workers:
+                os.kill(worker.pid, signum)
+        except Exception:
+            logger.exception('could not kill {}'.format(worker.pid))
+
+
+class AutoscalePool(WorkerPool):
+    '''
+    An extended pool implementation that automatically scales workers up and
+    down based on demand
+    '''
+
+    def __init__(self, *args, **kwargs):
+        self.max_workers = kwargs.pop('max_workers', None)
+        super(AutoscalePool, self).__init__(*args, **kwargs)
+
+        if self.max_workers is None:
+            settings_absmem = getattr(settings, 'SYSTEM_TASK_ABS_MEM', None)
+            if settings_absmem is not None:
+                total_memory_gb = int(settings_absmem)
+            else:
+                total_memory_gb = (psutil.virtual_memory().total >> 30) + 1  # noqa: round up
+            # 5 workers per GB of total memory
+            self.max_workers = (total_memory_gb * 5)
+
+    @property
+    def should_grow(self):
+        if len(self.workers) < self.min_workers:
+            # If we don't have at least min_workers, add more
+            return True
+        # If every worker is busy doing something, add more
+        return all([w.busy for w in self.workers])
+
+    @property
+    def full(self):
+        return len(self.workers) == self.max_workers
+
+    @property
+    def debug_meta(self):
+        return 'min={} max={}'.format(self.min_workers, self.max_workers)
+
+    def cleanup(self):
+        """
+        Perform some internal account and cleanup.  This is run on
+        every cluster node heartbeat:
+
+        1.  Discover worker processes that exited, and recover messages they
+            were handling.
+        2.  Clean up unnecessary, idle workers.
+        3.  Check to see if the database says this node is running any tasks
+            that aren't actually running.  If so, reap them.
+        """
+        orphaned = []
+        for w in self.workers[::]:
+            if not w.alive:
+                # the worker process has exited
+                # 1. take the task it was running and enqueue the error
+                #    callbacks
+                # 2. take any pending tasks delivered to its queue and
+                #    send them to another worker
+                logger.error('worker pid:{} is gone (exit={})'.format(w.pid, w.exitcode))
+                if w.current_task:
+                    if w.current_task != 'QUIT':
+                        try:
+                            for j in UnifiedJob.objects.filter(celery_task_id=w.current_task['uuid']):
+                                reaper.reap_job(j, 'failed')
+                        except Exception:
+                            logger.exception('failed to reap job UUID {}'.format(w.current_task['uuid']))
+                orphaned.extend(w.orphaned_tasks)
+                self.workers.remove(w)
+            elif w.idle and len(self.workers) > self.min_workers:
+                # the process has an empty queue (it's idle) and we have
+                # more processes in the pool than we need (> min)
+                # send this process a message so it will exit gracefully
+                # at the next opportunity
+                logger.warn('scaling down worker pid:{}'.format(w.pid))
+                w.quit()
+                self.workers.remove(w)
+
+        for m in orphaned:
+            # if all the workers are dead, spawn at least one
+            if not len(self.workers):
+                self.up()
+            idx = random.choice(range(len(self.workers)))
+            self.write(idx, m)
+
+        # if the database says a job is running on this node, but it's *not*,
+        # then reap it
+        running_uuids = []
+        for worker in self.workers:
+            worker.calculate_managed_tasks()
+            running_uuids.extend(worker.managed_tasks.keys())
+        try:
+            reaper.reap(excluded_uuids=running_uuids)
+        except Exception:
+            # we _probably_ failed here due to DB connectivity issues, so
+            # don't use our logger (it accesses the database for configuration)
+            _, _, tb = sys.exc_info()
+            traceback.print_tb(tb)
+
+    def up(self):
+        if self.full:
+            # if we can't spawn more workers, just toss this message into a
+            # random worker's backlog
+            idx = random.choice(range(len(self.workers)))
+            return idx, self.workers[idx]
+        else:
+            return super(AutoscalePool, self).up()
+
+    def write(self, preferred_queue, body):
+        try:
+            # when the cluster heartbeat occurs, clean up internally
+            if isinstance(body, dict) and 'cluster_node_heartbeat' in body['task']:
+                self.cleanup()
+            if self.should_grow:
+                self.up()
+            # we don't care about "preferred queue" round robin distribution, just
+            # find the first non-busy worker and claim it
+            workers = self.workers[:]
+            random.shuffle(workers)
+            for w in workers:
+                if not w.busy:
+                    w.put(body)
+                    break
+            else:
+                return super(AutoscalePool, self).write(preferred_queue, body)
+        except Exception:
+            for conn in connections.all():
+                # If the database connection has a hiccup, re-establish a new
+                # connection
+                conn.close_if_unusable_or_obsolete()
+            logger.exception('failed to write inbound message')

awx/main/dispatch/publish.py

@@ -0,0 +1,128 @@
+import inspect
+import logging
+import sys
+from uuid import uuid4
+
+from django.conf import settings
+from kombu import Connection, Exchange, Producer
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+def serialize_task(f):
+    return '.'.join([f.__module__, f.__name__])
+
+
+class task:
+    """
+    Used to decorate a function or class so that it can be run asynchronously
+    via the task dispatcher.  Tasks can be simple functions:
+
+    @task()
+    def add(a, b):
+        return a + b
+
+    ...or classes that define a `run` method:
+
+    @task()
+    class Adder:
+        def run(self, a, b):
+            return a + b
+
+    # Tasks can be run synchronously...
+    assert add(1, 1) == 2
+    assert Adder().run(1, 1) == 2
+
+    # ...or published to a queue:
+    add.apply_async([1, 1])
+    Adder.apply_async([1, 1])
+
+    # Tasks can also define a specific target queue or exchange type:
+
+    @task(queue='slow-tasks')
+    def snooze():
+        time.sleep(10)
+
+    @task(queue='tower_broadcast', exchange_type='fanout')
+    def announce():
+        print "Run this everywhere!"
+    """
+
+    def __init__(self, queue=None, exchange_type=None):
+        self.queue = queue
+        self.exchange_type = exchange_type
+
+    def __call__(self, fn=None):
+        queue = self.queue
+        exchange_type = self.exchange_type
+
+        class PublisherMixin(object):
+
+            queue = None
+
+            @classmethod
+            def delay(cls, *args, **kwargs):
+                return cls.apply_async(args, kwargs)
+
+            @classmethod
+            def apply_async(cls, args=None, kwargs=None, queue=None, uuid=None, **kw):
+                task_id = uuid or str(uuid4())
+                args = args or []
+                kwargs = kwargs or {}
+                queue = (
+                    queue or
+                    getattr(cls.queue, 'im_func', cls.queue) or
+                    settings.CELERY_DEFAULT_QUEUE
+                )
+                obj = {
+                    'uuid': task_id,
+                    'args': args,
+                    'kwargs': kwargs,
+                    'task': cls.name
+                }
+                obj.update(**kw)
+                if callable(queue):
+                    queue = queue()
+                if not settings.IS_TESTING(sys.argv):
+                    with Connection(settings.BROKER_URL) as conn:
+                        exchange = Exchange(queue, type=exchange_type or 'direct')
+                        producer = Producer(conn)
+                        logger.debug('publish {}({}, queue={})'.format(
+                            cls.name,
+                            task_id,
+                            queue
+                        ))
+                        producer.publish(obj,
+                                         serializer='json',
+                                         compression='bzip2',
+                                         exchange=exchange,
+                                         declare=[exchange],
+                                         delivery_mode="persistent",
+                                         routing_key=queue)
+                return (obj, queue)
+
+        # If the object we're wrapping *is* a class (e.g., RunJob), return
+        # a *new* class that inherits from the wrapped class *and* BaseTask
+        # In this way, the new class returned by our decorator is the class
+        # being decorated *plus* PublisherMixin so cls.apply_async() and
+        # cls.delay() work
+        bases = []
+        ns = {'name': serialize_task(fn), 'queue': queue}
+        if inspect.isclass(fn):
+            bases = list(fn.__bases__)
+            ns.update(fn.__dict__)
+        cls = type(
+            fn.__name__,
+            tuple(bases + [PublisherMixin]),
+            ns
+        )
+        if inspect.isclass(fn):
+            return cls
+
+        # if the object being decorated is *not* a class (it's a Python
+        # function), make fn.apply_async and fn.delay proxy through to the
+        # PublisherMixin we dynamically created above
+        setattr(fn, 'name', cls.name)
+        setattr(fn, 'apply_async', cls.apply_async)
+        setattr(fn, 'delay', cls.delay)
+        return fn

awx/main/dispatch/reaper.py

@@ -0,0 +1,46 @@
+from datetime import timedelta
+import logging
+
+from django.db.models import Q
+from django.utils.timezone import now as tz_now
+from django.contrib.contenttypes.models import ContentType
+
+from awx.main.models import Instance, UnifiedJob, WorkflowJob
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+def reap_job(j, status):
+    j.status = status
+    j.start_args = ''  # blank field to remove encrypted passwords
+    j.job_explanation += ' '.join((
+        'Task was marked as running in Tower but was not present in',
+        'the job queue, so it has been marked as failed.',
+    ))
+    j.save(update_fields=['status', 'start_args', 'job_explanation'])
+    if hasattr(j, 'send_notification_templates'):
+        j.send_notification_templates('failed')
+    j.websocket_emit_status(status)
+    logger.error(
+        '{} is no longer running; reaping'.format(j.log_format)
+    )
+
+
+def reap(instance=None, status='failed', excluded_uuids=[]):
+    '''
+    Reap all jobs in waiting|running for this instance.
+    '''
+    me = instance or Instance.objects.me()
+    now = tz_now()
+    workflow_ctype_id = ContentType.objects.get_for_model(WorkflowJob).id
+    jobs = UnifiedJob.objects.filter(
+        (
+            Q(status='running') |
+            Q(status='waiting', modified__lte=now - timedelta(seconds=60))
+        ) & (
+            Q(execution_node=me.hostname) |
+            Q(controller_node=me.hostname)
+        ) & ~Q(polymorphic_ctype_id=workflow_ctype_id)
+    ).exclude(celery_task_id__in=excluded_uuids)
+    for j in jobs:
+        reap_job(j, status)

awx/main/dispatch/worker/__init__.py

@@ -0,0 +1,3 @@
+from .base import AWXConsumer, BaseWorker  # noqa
+from .callback import CallbackBrokerWorker  # noqa
+from .task import TaskWorker  # noqa

awx/main/dispatch/worker/base.py

@@ -0,0 +1,146 @@
+# Copyright (c) 2018 Ansible by Red Hat
+# All Rights Reserved.
+
+import os
+import logging
+import signal
+from uuid import UUID
+from Queue import Empty as QueueEmpty
+
+from kombu import Producer
+from kombu.mixins import ConsumerMixin
+
+from awx.main.dispatch.pool import WorkerPool
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+def signame(sig):
+    return dict(
+        (k, v) for v, k in signal.__dict__.items()
+        if v.startswith('SIG') and not v.startswith('SIG_')
+    )[sig]
+
+
+class WorkerSignalHandler:
+
+    def __init__(self):
+        self.kill_now = False
+        signal.signal(signal.SIGINT, self.exit_gracefully)
+
+    def exit_gracefully(self, *args, **kwargs):
+        self.kill_now = True
+
+
+class AWXConsumer(ConsumerMixin):
+
+    def __init__(self, name, connection, worker, queues=[], pool=None):
+        self.connection = connection
+        self.total_messages = 0
+        self.queues = queues
+        self.worker = worker
+        self.pool = pool
+        if pool is None:
+            self.pool = WorkerPool()
+        self.pool.init_workers(self.worker.work_loop)
+
+    def get_consumers(self, Consumer, channel):
+        logger.debug(self.listening_on)
+        return [Consumer(queues=self.queues, accept=['json'],
+                         callbacks=[self.process_task])]
+
+    @property
+    def listening_on(self):
+        return 'listening on {}'.format([
+            '{} [{}]'.format(q.name, q.exchange.type) for q in self.queues
+        ])
+
+    def control(self, body, message):
+        logger.warn(body)
+        control = body.get('control')
+        if control in ('status', 'running'):
+            producer = Producer(
+                channel=self.connection,
+                routing_key=message.properties['reply_to']
+            )
+            if control == 'status':
+                msg = '\n'.join([self.listening_on, self.pool.debug()])
+            elif control == 'running':
+                msg = []
+                for worker in self.pool.workers:
+                    worker.calculate_managed_tasks()
+                    msg.extend(worker.managed_tasks.keys())
+            producer.publish(msg)
+        elif control == 'reload':
+            for worker in self.pool.workers:
+                worker.quit()
+        else:
+            logger.error('unrecognized control message: {}'.format(control))
+        message.ack()
+
+    def process_task(self, body, message):
+        if 'control' in body:
+            return self.control(body, message)
+        if len(self.pool):
+            if "uuid" in body and body['uuid']:
+                try:
+                    queue = UUID(body['uuid']).int % len(self.pool)
+                except Exception:
+                    queue = self.total_messages % len(self.pool)
+            else:
+                queue = self.total_messages % len(self.pool)
+        else:
+            queue = 0
+        self.pool.write(queue, body)
+        self.total_messages += 1
+        message.ack()
+
+    def run(self, *args, **kwargs):
+        signal.signal(signal.SIGINT, self.stop)
+        signal.signal(signal.SIGTERM, self.stop)
+        self.worker.on_start()
+        super(AWXConsumer, self).run(*args, **kwargs)
+
+    def stop(self, signum, frame):
+        self.should_stop = True  # this makes the kombu mixin stop consuming
+        logger.debug('received {}, stopping'.format(signame(signum)))
+        self.worker.on_stop()
+        raise SystemExit()
+
+
+class BaseWorker(object):
+
+    def work_loop(self, queue, finished, idx, *args):
+        ppid = os.getppid()
+        signal_handler = WorkerSignalHandler()
+        while not signal_handler.kill_now:
+            # if the parent PID changes, this process has been orphaned
+            # via e.g., segfault or sigkill, we should exit too
+            if os.getppid() != ppid:
+                break
+            try:
+                body = queue.get(block=True, timeout=1)
+                if body == 'QUIT':
+                    break
+            except QueueEmpty:
+                continue
+            except Exception as e:
+                logger.error("Exception on worker {}, restarting: ".format(idx) + str(e))
+                continue
+            try:
+                self.perform_work(body, *args)
+            finally:
+                if 'uuid' in body:
+                    uuid = body['uuid']
+                    logger.debug('task {} is finished'.format(uuid))
+                    finished.put(uuid)
+        logger.warn('worker exiting gracefully pid:{}'.format(os.getpid()))
+
+    def perform_work(self, body):
+        raise NotImplementedError()
+
+    def on_start(self):
+        pass
+
+    def on_stop(self):
+        pass

awx/main/dispatch/worker/callback.py

@@ -0,0 +1,130 @@
+import logging
+import time
+import os
+import signal
+import traceback
+
+from django.conf import settings
+from django.db import DatabaseError, OperationalError, connection as django_connection
+from django.db.utils import InterfaceError, InternalError
+
+from awx.main.consumers import emit_channel_notification
+from awx.main.models import (JobEvent, AdHocCommandEvent, ProjectUpdateEvent,
+                             InventoryUpdateEvent, SystemJobEvent, UnifiedJob)
+
+from .base import BaseWorker
+
+logger = logging.getLogger('awx.main.commands.run_callback_receiver')
+
+
+class CallbackBrokerWorker(BaseWorker):
+    '''
+    A worker implementation that deserializes callback event data and persists
+    it into the database.
+
+    The code that *builds* these types of messages is found in the AWX display
+    callback (`awx.lib.awx_display_callback`).
+    '''
+
+    MAX_RETRIES = 2
+
+    def perform_work(self, body):
+        try:
+            event_map = {
+                'job_id': JobEvent,
+                'ad_hoc_command_id': AdHocCommandEvent,
+                'project_update_id': ProjectUpdateEvent,
+                'inventory_update_id': InventoryUpdateEvent,
+                'system_job_id': SystemJobEvent,
+            }
+
+            if not any([key in body for key in event_map]):
+                raise Exception('Payload does not have a job identifier')
+
+            def _save_event_data():
+                for key, cls in event_map.items():
+                    if key in body:
+                        cls.create_from_data(**body)
+
+            job_identifier = 'unknown job'
+            job_key = 'unknown'
+            for key in event_map.keys():
+                if key in body:
+                    job_identifier = body[key]
+                    job_key = key
+                    break
+
+            if settings.DEBUG:
+                from pygments import highlight
+                from pygments.lexers import PythonLexer
+                from pygments.formatters import Terminal256Formatter
+                from pprint import pformat
+                if body.get('event') == 'EOF':
+                    event_thing = 'EOF event'
+                else:
+                    event_thing = 'event {}'.format(body.get('counter', 'unknown'))
+                logger.info('Callback worker received {} for {} {}'.format(
+                    event_thing, job_key[:-len('_id')], job_identifier
+                ))
+                logger.debug('Body: {}'.format(
+                    highlight(pformat(body, width=160), PythonLexer(), Terminal256Formatter(style='friendly'))
+                )[:1024 * 4])
+
+            if body.get('event') == 'EOF':
+                try:
+                    final_counter = body.get('final_counter', 0)
+                    logger.info('Event processing is finished for Job {}, sending notifications'.format(job_identifier))
+                    # EOF events are sent when stdout for the running task is
+                    # closed. don't actually persist them to the database; we
+                    # just use them to report `summary` websocket events as an
+                    # approximation for when a job is "done"
+                    emit_channel_notification(
+                        'jobs-summary',
+                        dict(group_name='jobs', unified_job_id=job_identifier, final_counter=final_counter)
+                    )
+                    # Additionally, when we've processed all events, we should
+                    # have all the data we need to send out success/failure
+                    # notification templates
+                    uj = UnifiedJob.objects.get(pk=job_identifier)
+                    if hasattr(uj, 'send_notification_templates'):
+                        retries = 0
+                        while retries < 5:
+                            if uj.finished:
+                                uj.send_notification_templates('succeeded' if uj.status == 'successful' else 'failed')
+                                break
+                            else:
+                                # wait a few seconds to avoid a race where the
+                                # events are persisted _before_ the UJ.status
+                                # changes from running -> successful
+                                retries += 1
+                                time.sleep(1)
+                                uj = UnifiedJob.objects.get(pk=job_identifier)
+                except Exception:
+                    logger.exception('Worker failed to emit notifications: Job {}'.format(job_identifier))
+                return
+
+            retries = 0
+            while retries <= self.MAX_RETRIES:
+                try:
+                    _save_event_data()
+                    break
+                except (OperationalError, InterfaceError, InternalError):
+                    if retries >= self.MAX_RETRIES:
+                        logger.exception('Worker could not re-establish database connectivity, shutting down gracefully: Job {}'.format(job_identifier))
+                        os.kill(os.getppid(), signal.SIGINT)
+                        return
+                    delay = 60 * retries
+                    logger.exception('Database Error Saving Job Event, retry #{i} in {delay} seconds:'.format(
+                        i=retries + 1,
+                        delay=delay
+                    ))
+                    django_connection.close()
+                    time.sleep(delay)
+                    retries += 1
+                except DatabaseError:
+                    logger.exception('Database Error Saving Job Event for Job {}'.format(job_identifier))
+                    break
+        except Exception as exc:
+            tb = traceback.format_exc()
+            logger.error('Callback Task Processor Raised Exception: %r', exc)
+            logger.error('Detail: {}'.format(tb))

awx/main/dispatch/worker/task.py

@@ -0,0 +1,118 @@
+import inspect
+import logging
+import importlib
+import sys
+import traceback
+
+import six
+from django import db
+
+from awx.main.tasks import dispatch_startup, inform_cluster_of_shutdown
+
+from .base import BaseWorker
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+class TaskWorker(BaseWorker):
+    '''
+    A worker implementation that deserializes task messages and runs native
+    Python code.
+
+    The code that *builds* these types of messages is found in
+    `awx.main.dispatch.publish`.
+    '''
+
+    @classmethod
+    def resolve_callable(cls, task):
+        '''
+        Transform a dotted notation task into an imported, callable function, e.g.,
+
+        awx.main.tasks.delete_inventory
+        awx.main.tasks.RunProjectUpdate
+        '''
+        module, target = task.rsplit('.', 1)
+        module = importlib.import_module(module)
+        _call = None
+        if hasattr(module, target):
+            _call = getattr(module, target, None)
+        return _call
+
+    def run_callable(self, body):
+        '''
+        Given some AMQP message, import the correct Python code and run it.
+        '''
+        task = body['task']
+        uuid = body.get('uuid', '<unknown>')
+        args = body.get('args', [])
+        kwargs = body.get('kwargs', {})
+        _call = TaskWorker.resolve_callable(task)
+        if inspect.isclass(_call):
+            # the callable is a class, e.g., RunJob; instantiate and
+            # return its `run()` method
+            _call = _call().run
+        # don't print kwargs, they often contain launch-time secrets
+        logger.debug('task {} starting {}(*{})'.format(uuid, task, args))
+        return _call(*args, **kwargs)
+
+    def perform_work(self, body):
+        '''
+        Import and run code for a task e.g.,
+
+        body = {
+            'args': [8],
+            'callbacks': [{
+                'args': [],
+                'kwargs': {}
+                'task': u'awx.main.tasks.handle_work_success'
+            }],
+            'errbacks': [{
+                'args': [],
+                'kwargs': {},
+                'task': 'awx.main.tasks.handle_work_error'
+            }],
+            'kwargs': {},
+            'task': u'awx.main.tasks.RunProjectUpdate'
+        }
+        '''
+        for conn in db.connections.all():
+            # If the database connection has a hiccup during at task, close it
+            # so we can establish a new connection
+            conn.close_if_unusable_or_obsolete()
+        result = None
+        try:
+            result = self.run_callable(body)
+        except Exception as exc:
+
+            try:
+                if getattr(exc, 'is_awx_task_error', False):
+                    # Error caused by user / tracked in job output
+                    logger.warning(six.text_type("{}").format(exc))
+                else:
+                    task = body['task']
+                    args = body.get('args', [])
+                    kwargs = body.get('kwargs', {})
+                    logger.exception('Worker failed to run task {}(*{}, **{}'.format(
+                        task, args, kwargs
+                    ))
+            except Exception:
+                # It's fairly critical that this code _not_ raise exceptions on logging
+                # If you configure external logging in a way that _it_ fails, there's
+                # not a lot we can do here; sys.stderr.write is a final hail mary
+                _, _, tb = sys.exc_info()
+                traceback.print_tb(tb)
+
+            for callback in body.get('errbacks', []) or []:
+                callback['uuid'] = body['uuid']
+                self.perform_work(callback)
+
+        for callback in body.get('callbacks', []) or []:
+            callback['uuid'] = body['uuid']
+            self.perform_work(callback)
+        return result
+
+    def on_start(self):
+        dispatch_startup()
+
+    def on_stop(self):
+        inform_cluster_of_shutdown()

awx/main/exceptions.py

@@ -4,11 +4,6 @@
 import six
 
 
-# Celery does not respect exception type when using a serializer different than pickle;
-# and awx uses the json serializer
-# https://github.com/celery/celery/issues/3586
-
-
 class _AwxTaskError():
     def build_exception(self, task, message=None):
         if message is None:
@@ -36,5 +31,3 @@ class _AwxTaskError():
 
 
 AwxTaskError = _AwxTaskError()
-
-

awx/main/expect/isolated_manager.py

@@ -38,7 +38,7 @@ class IsolatedManager(object):
         :param stdout_handle:       a file-like object for capturing stdout
         :param ssh_key_path:        a filepath where SSH key data can be read
         :param expect_passwords:    a dict of regular expression password prompts
-                                    to input values, i.e., {r'Password:\s*?$':
+                                    to input values, i.e., {r'Password:*?$':
                                     'some_password'}
         :param cancelled_callback:  a callable - which returns `True` or `False`
                                     - signifying if the job has been prematurely

awx/main/expect/run.py

@@ -71,7 +71,7 @@ def run_pexpect(args, cwd, env, logfile,
                                 - signifying if the job has been prematurely
                                   cancelled
     :param expect_passwords:    a dict of regular expression password prompts
-                                to input values, i.e., {r'Password:\s*?$':
+                                to input values, i.e., {r'Password:*?$':
                                 'some_password'}
     :param extra_update_fields: a dict used to specify DB fields which should
                                 be updated on the underlying model

awx/main/fields.py

@@ -46,7 +46,7 @@ from awx.main.utils.filters import SmartFilter
 from awx.main.utils.encryption import encrypt_value, decrypt_value, get_encryption_key
 from awx.main.validators import validate_ssh_private_key
 from awx.main.models.rbac import batch_role_ancestor_rebuilding, Role
-from awx.main.constants import CHOICES_PRIVILEGE_ESCALATION_METHODS
+from awx.main.constants import CHOICES_PRIVILEGE_ESCALATION_METHODS, ENV_BLACKLIST
 from awx.main import utils
 
 
@@ -573,10 +573,10 @@ class CredentialInputField(JSONSchemaField):
                 # string)
                 match = re.search(
                     # 'foo' is a dependency of 'bar'
-                    "'"         # apostrophe
-                    "([^']+)"   # one or more non-apostrophes (first group)
-                    "'[\w ]+'"  # one or more words/spaces
-                    "([^']+)",  # second group
+                    r"'"         # apostrophe
+                    r"([^']+)"   # one or more non-apostrophes (first group)
+                    r"'[\w ]+'"  # one or more words/spaces
+                    r"([^']+)",  # second group
                     error.message,
                 )
                 if match:
@@ -755,7 +755,7 @@ class CredentialTypeInjectorField(JSONSchemaField):
                 'file': {
                     'type': 'object',
                     'patternProperties': {
-                        '^template(\.[a-zA-Z_]+[a-zA-Z0-9_]*)?$': {'type': 'string'},
+                        r'^template(\.[a-zA-Z_]+[a-zA-Z0-9_]*)?$': {'type': 'string'},
                     },
                     'additionalProperties': False,
                 },
@@ -767,7 +767,12 @@ class CredentialTypeInjectorField(JSONSchemaField):
                         # of underscores, digits, and alphabetics from the portable
                         # character set. The first character of a name is not
                         # a digit.
-                        '^[a-zA-Z_]+[a-zA-Z0-9_]*$': {'type': 'string'},
+                        '^[a-zA-Z_]+[a-zA-Z0-9_]*$': {
+                            'type': 'string',
+                            # The environment variable _value_ can be any ascii,
+                            # but pexpect will choke on any unicode
+                            'pattern': '^[\x00-\x7F]*$'
+                        },
                     },
                     'additionalProperties': False,
                 },
@@ -783,6 +788,19 @@ class CredentialTypeInjectorField(JSONSchemaField):
             'additionalProperties': False
         }
 
+    def validate_env_var_allowed(self, env_var):
+        if env_var.startswith('ANSIBLE_'):
+            raise django_exceptions.ValidationError(
+                _('Environment variable {} may affect Ansible configuration so its '
+                  'use is not allowed in credentials.').format(env_var),
+                code='invalid', params={'value': env_var},
+            )
+        if env_var in ENV_BLACKLIST:
+            raise django_exceptions.ValidationError(
+                _('Environment variable {} is blacklisted from use in credentials.').format(env_var),
+                code='invalid', params={'value': env_var},
+            )
+
     def validate(self, value, model_instance):
         super(CredentialTypeInjectorField, self).validate(
             value, model_instance
@@ -834,6 +852,9 @@ class CredentialTypeInjectorField(JSONSchemaField):
                 setattr(valid_namespace['tower'].filename, template_name, 'EXAMPLE_FILENAME')
 
         for type_, injector in value.items():
+            if type_ == 'env':
+                for key in injector.keys():
+                    self.validate_env_var_allowed(key)
             for key, tmpl in injector.items():
                 try:
                     Environment(

awx/main/management/commands/deprovision_instance.py

@@ -41,14 +41,18 @@ class Command(BaseCommand):
         with advisory_lock('instance_registration_%s' % hostname):
             instance = Instance.objects.filter(hostname=hostname)
             if instance.exists():
+                isolated = instance.first().is_isolated()
                 instance.delete()
                 print("Instance Removed")
-                result = subprocess.Popen("rabbitmqctl forget_cluster_node rabbitmq@{}".format(hostname), shell=True).wait()
-                if result != 0:
-                    print("Node deprovisioning may have failed when attempting to "
-                          "remove the RabbitMQ instance {} from the cluster".format(hostname))
-                else:
+                if isolated:
                     print('Successfully deprovisioned {}'.format(hostname))
+                else:
+                    result = subprocess.Popen("rabbitmqctl forget_cluster_node rabbitmq@{}".format(hostname), shell=True).wait()
+                    if result != 0:
+                        print("Node deprovisioning may have failed when attempting to "
+                              "remove the RabbitMQ instance {} from the cluster".format(hostname))
+                    else:
+                        print('Successfully deprovisioned {}'.format(hostname))
                 print('(changed: True)')
             else:
                 print('No instance found matching name {}'.format(hostname))

awx/main/management/commands/inventory_import.py

@@ -938,7 +938,7 @@ class Command(BaseCommand):
         self.exclude_empty_groups = bool(options.get('exclude_empty_groups', False))
         self.instance_id_var = options.get('instance_id_var', None)
 
-        self.celery_invoked = False if os.getenv('INVENTORY_SOURCE_ID', None) is None else True
+        self.invoked_from_dispatcher = False if os.getenv('INVENTORY_SOURCE_ID', None) is None else True
 
         # Load inventory and related objects from database.
         if self.inventory_name and self.inventory_id:
@@ -1062,7 +1062,7 @@ class Command(BaseCommand):
                 exc = e
             transaction.rollback()
 
-        if self.celery_invoked is False:
+        if self.invoked_from_dispatcher is False:
             with ignore_inventory_computed_fields():
                 self.inventory_update = InventoryUpdate.objects.get(pk=self.inventory_update.pk)
                 self.inventory_update.result_traceback = tb

awx/main/management/commands/run_callback_receiver.py

@@ -1,231 +1,11 @@
 # Copyright (c) 2015 Ansible, Inc.
 # All Rights Reserved.
 
-# Python
-import logging
-import os
-import signal
-import time
-from uuid import UUID
-from multiprocessing import Process
-from multiprocessing import Queue as MPQueue
-from Queue import Empty as QueueEmpty
-from Queue import Full as QueueFull
-
-from kombu import Connection, Exchange, Queue
-from kombu.mixins import ConsumerMixin
-
-# Django
 from django.conf import settings
 from django.core.management.base import BaseCommand
-from django.db import connection as django_connection
-from django.db import DatabaseError, OperationalError
-from django.db.utils import InterfaceError, InternalError
-from django.core.cache import cache as django_cache
+from kombu import Connection, Exchange, Queue
 
-# AWX
-from awx.main.models import * # noqa
-from awx.main.consumers import emit_channel_notification
-
-logger = logging.getLogger('awx.main.commands.run_callback_receiver')
-
-
-class WorkerSignalHandler:
-
-    def __init__(self):
-        self.kill_now = False
-        signal.signal(signal.SIGINT, self.exit_gracefully)
-        signal.signal(signal.SIGTERM, self.exit_gracefully)
-
-    def exit_gracefully(self, *args, **kwargs):
-        self.kill_now = True
-
-
-class CallbackBrokerWorker(ConsumerMixin):
-
-    MAX_RETRIES = 2
-
-    def __init__(self, connection, use_workers=True):
-        self.connection = connection
-        self.worker_queues = []
-        self.total_messages = 0
-        self.init_workers(use_workers)
-
-    def init_workers(self, use_workers=True):
-        def shutdown_handler(active_workers):
-            def _handler(signum, frame):
-                try:
-                    for active_worker in active_workers:
-                        active_worker.terminate()
-                    signal.signal(signum, signal.SIG_DFL)
-                    os.kill(os.getpid(), signum) # Rethrow signal, this time without catching it
-                except Exception:
-                    logger.exception('Error in shutdown_handler')
-            return _handler
-
-        if use_workers:
-            for idx in range(settings.JOB_EVENT_WORKERS):
-                queue_actual = MPQueue(settings.JOB_EVENT_MAX_QUEUE_SIZE)
-                w = Process(target=self.callback_worker, args=(queue_actual, idx,))
-                if settings.DEBUG:
-                    logger.info('Starting worker %s' % str(idx))
-                self.worker_queues.append([0, queue_actual, w])
-
-            # It's important to close these _right before_ we fork; we
-            # don't want the forked processes to inherit the open sockets
-            # for the DB and memcached connections (that way lies race
-            # conditions)
-            django_connection.close()
-            django_cache.close()
-            for _, _, w in self.worker_queues:
-                w.start()
-
-        elif settings.DEBUG:
-            logger.warn('Started callback receiver (no workers)')
-
-        signal.signal(signal.SIGINT, shutdown_handler([p[2] for p in self.worker_queues]))
-        signal.signal(signal.SIGTERM, shutdown_handler([p[2] for p in self.worker_queues]))
-
-    def get_consumers(self, Consumer, channel):
-        return [Consumer(queues=[Queue(settings.CALLBACK_QUEUE,
-                                       Exchange(settings.CALLBACK_QUEUE, type='direct'),
-                                       routing_key=settings.CALLBACK_QUEUE)],
-                         accept=['json'],
-                         callbacks=[self.process_task])]
-
-    def process_task(self, body, message):
-        if "uuid" in body and body['uuid']:
-            try:
-                queue = UUID(body['uuid']).int % settings.JOB_EVENT_WORKERS
-            except Exception:
-                queue = self.total_messages % settings.JOB_EVENT_WORKERS
-        else:
-            queue = self.total_messages % settings.JOB_EVENT_WORKERS
-        self.write_queue_worker(queue, body)
-        self.total_messages += 1
-        message.ack()
-
-    def write_queue_worker(self, preferred_queue, body):
-        queue_order = sorted(range(settings.JOB_EVENT_WORKERS), cmp=lambda x, y: -1 if x==preferred_queue else 0)
-        write_attempt_order = []
-        for queue_actual in queue_order:
-            try:
-                worker_actual = self.worker_queues[queue_actual]
-                worker_actual[1].put(body, block=True, timeout=5)
-                worker_actual[0] += 1
-                return queue_actual
-            except QueueFull:
-                pass
-            except Exception:
-                import traceback
-                tb = traceback.format_exc()
-                logger.warn("Could not write to queue %s" % preferred_queue)
-                logger.warn("Detail: {}".format(tb))
-            write_attempt_order.append(preferred_queue)
-        logger.warn("Could not write payload to any queue, attempted order: {}".format(write_attempt_order))
-        return None
-
-    def callback_worker(self, queue_actual, idx):
-        signal_handler = WorkerSignalHandler()
-        while not signal_handler.kill_now:
-            try:
-                body = queue_actual.get(block=True, timeout=1)
-            except QueueEmpty:
-                continue
-            except Exception as e:
-                logger.error("Exception on worker thread, restarting: " + str(e))
-                continue
-            try:
-
-                event_map = {
-                    'job_id': JobEvent,
-                    'ad_hoc_command_id': AdHocCommandEvent,
-                    'project_update_id': ProjectUpdateEvent,
-                    'inventory_update_id': InventoryUpdateEvent,
-                    'system_job_id': SystemJobEvent,
-                }
-
-                if not any([key in body for key in event_map]):
-                    raise Exception('Payload does not have a job identifier')
-                if settings.DEBUG:
-                    from pygments import highlight
-                    from pygments.lexers import PythonLexer
-                    from pygments.formatters import Terminal256Formatter
-                    from pprint import pformat
-                    logger.info('Body: {}'.format(
-                        highlight(pformat(body, width=160), PythonLexer(), Terminal256Formatter(style='friendly'))
-                    )[:1024 * 4])
-
-                def _save_event_data():
-                    for key, cls in event_map.items():
-                        if key in body:
-                            cls.create_from_data(**body)
-
-                job_identifier = 'unknown job'
-                for key in event_map.keys():
-                    if key in body:
-                        job_identifier = body[key]
-                        break
-
-                if body.get('event') == 'EOF':
-                    try:
-                        final_counter = body.get('final_counter', 0)
-                        logger.info('Event processing is finished for Job {}, sending notifications'.format(job_identifier))
-                        # EOF events are sent when stdout for the running task is
-                        # closed. don't actually persist them to the database; we
-                        # just use them to report `summary` websocket events as an
-                        # approximation for when a job is "done"
-                        emit_channel_notification(
-                            'jobs-summary',
-                            dict(group_name='jobs', unified_job_id=job_identifier, final_counter=final_counter)
-                        )
-                        # Additionally, when we've processed all events, we should
-                        # have all the data we need to send out success/failure
-                        # notification templates
-                        uj = UnifiedJob.objects.get(pk=job_identifier)
-                        if hasattr(uj, 'send_notification_templates'):
-                            retries = 0
-                            while retries < 5:
-                                if uj.finished:
-                                    uj.send_notification_templates('succeeded' if uj.status == 'successful' else 'failed')
-                                    break
-                                else:
-                                    # wait a few seconds to avoid a race where the
-                                    # events are persisted _before_ the UJ.status
-                                    # changes from running -> successful
-                                    retries += 1
-                                    time.sleep(1)
-                                    uj = UnifiedJob.objects.get(pk=job_identifier)
-                    except Exception:
-                        logger.exception('Worker failed to emit notifications: Job {}'.format(job_identifier))
-                    continue
-
-                retries = 0
-                while retries <= self.MAX_RETRIES:
-                    try:
-                        _save_event_data()
-                        break
-                    except (OperationalError, InterfaceError, InternalError) as e:
-                        if retries >= self.MAX_RETRIES:
-                            logger.exception('Worker could not re-establish database connectivity, shutting down gracefully: Job {}'.format(job_identifier))
-                            os.kill(os.getppid(), signal.SIGINT)
-                            return
-                        delay = 60 * retries
-                        logger.exception('Database Error Saving Job Event, retry #{i} in {delay} seconds:'.format(
-                            i=retries + 1,
-                            delay=delay
-                        ))
-                        django_connection.close()
-                        time.sleep(delay)
-                        retries += 1
-                    except DatabaseError as e:
-                        logger.exception('Database Error Saving Job Event for Job {}'.format(job_identifier))
-                        break
-            except Exception as exc:
-                import traceback
-                tb = traceback.format_exc()
-                logger.error('Callback Task Processor Raised Exception: %r', exc)
-                logger.error('Detail: {}'.format(tb))
+from awx.main.dispatch.worker import AWXConsumer, CallbackBrokerWorker
 
 
 class Command(BaseCommand):
@@ -238,8 +18,22 @@ class Command(BaseCommand):
 
     def handle(self, *arg, **options):
         with Connection(settings.BROKER_URL) as conn:
+            consumer = None
             try:
-                worker = CallbackBrokerWorker(conn)
-                worker.run()
+                consumer = AWXConsumer(
+                    'callback_receiver',
+                    conn,
+                    CallbackBrokerWorker(),
+                    [
+                        Queue(
+                            settings.CALLBACK_QUEUE,
+                            Exchange(settings.CALLBACK_QUEUE, type='direct'),
+                            routing_key=settings.CALLBACK_QUEUE
+                        )
+                    ]
+                )
+                consumer.run()
             except KeyboardInterrupt:
                 print('Terminating Callback Receiver')
+                if consumer:
+                    consumer.stop()

awx/main/management/commands/run_dispatcher.py

@@ -0,0 +1,128 @@
+# Copyright (c) 2015 Ansible, Inc.
+# All Rights Reserved.
+import os
+import logging
+from multiprocessing import Process
+
+from django.conf import settings
+from django.core.cache import cache as django_cache
+from django.core.management.base import BaseCommand
+from django.db import connection as django_connection, connections
+from kombu import Connection, Exchange, Queue
+
+from awx.main.dispatch import get_local_queuename, reaper
+from awx.main.dispatch.control import Control
+from awx.main.dispatch.pool import AutoscalePool
+from awx.main.dispatch.worker import AWXConsumer, TaskWorker
+
+logger = logging.getLogger('awx.main.dispatch')
+
+
+def construct_bcast_queue_name(common_name):
+    return common_name.encode('utf8') + '_' + settings.CLUSTER_HOST_ID
+
+
+class Command(BaseCommand):
+    help = 'Launch the task dispatcher'
+
+    def add_arguments(self, parser):
+        parser.add_argument('--status', dest='status', action='store_true',
+                            help='print the internal state of any running dispatchers')
+        parser.add_argument('--running', dest='running', action='store_true',
+                            help='print the UUIDs of any tasked managed by this dispatcher')
+        parser.add_argument('--reload', dest='reload', action='store_true',
+                            help=('cause the dispatcher to recycle all of its worker processes;'
+                                  'running jobs will run to completion first'))
+
+    def beat(self):
+        from celery import Celery
+        from celery.beat import PersistentScheduler
+        from celery.apps import beat
+
+        class AWXScheduler(PersistentScheduler):
+
+            def __init__(self, *args, **kwargs):
+                self.ppid = os.getppid()
+                super(AWXScheduler, self).__init__(*args, **kwargs)
+
+            def setup_schedule(self):
+                super(AWXScheduler, self).setup_schedule()
+                self.update_from_dict(settings.CELERYBEAT_SCHEDULE)
+
+            def tick(self, *args, **kwargs):
+                if os.getppid() != self.ppid:
+                    # if the parent PID changes, this process has been orphaned
+                    # via e.g., segfault or sigkill, we should exit too
+                    raise SystemExit()
+                return super(AWXScheduler, self).tick(*args, **kwargs)
+
+            def apply_async(self, entry, producer=None, advance=True, **kwargs):
+                for conn in connections.all():
+                    # If the database connection has a hiccup, re-establish a new
+                    # connection
+                    conn.close_if_unusable_or_obsolete()
+                task = TaskWorker.resolve_callable(entry.task)
+                result, queue = task.apply_async()
+
+                class TaskResult(object):
+                    id = result['uuid']
+
+                return TaskResult()
+
+        app = Celery()
+        app.conf.BROKER_URL = settings.BROKER_URL
+        app.conf.CELERY_TASK_RESULT_EXPIRES = False
+        beat.Beat(
+            30,
+            app,
+            schedule='/var/lib/awx/beat.db', scheduler_cls=AWXScheduler
+        ).run()
+
+    def handle(self, *arg, **options):
+        if options.get('status'):
+            print Control('dispatcher').status()
+            return
+        if options.get('running'):
+            print Control('dispatcher').running()
+            return
+        if options.get('reload'):
+            return Control('dispatcher').control({'control': 'reload'})
+
+        # It's important to close these because we're _about_ to fork, and we
+        # don't want the forked processes to inherit the open sockets
+        # for the DB and memcached connections (that way lies race conditions)
+        django_connection.close()
+        django_cache.close()
+        beat = Process(target=self.beat)
+        beat.daemon = True
+        beat.start()
+
+        reaper.reap()
+        consumer = None
+        with Connection(settings.BROKER_URL) as conn:
+            try:
+                bcast = 'tower_broadcast_all'
+                queues = [
+                    Queue(q, Exchange(q), routing_key=q)
+                    for q in (settings.AWX_CELERY_QUEUES_STATIC + [get_local_queuename()])
+                ]
+                queues.append(
+                    Queue(
+                        construct_bcast_queue_name(bcast),
+                        exchange=Exchange(bcast, type='fanout'),
+                        routing_key=bcast,
+                        reply=True
+                    )
+                )
+                consumer = AWXConsumer(
+                    'dispatcher',
+                    conn,
+                    TaskWorker(),
+                    queues,
+                    AutoscalePool(min_workers=4)
+                )
+                consumer.run()
+            except KeyboardInterrupt:
+                logger.debug('Terminating Task Dispatcher')
+                if consumer:
+                    consumer.stop()

awx/main/management/commands/watch_celery.py

@@ -1,66 +0,0 @@
-import datetime
-import os
-import signal
-import subprocess
-import sys
-import time
-
-from celery import Celery
-from django.core.management.base import BaseCommand
-from django.conf import settings
-
-
-class Command(BaseCommand):
-    """Watch local celery workers"""
-    help=("Sends a periodic ping to the local celery process over AMQP to ensure "
-          "it's responsive; this command is only intended to run in an environment "
-          "where celeryd is running")
-
-    #
-    # Just because celery is _running_ doesn't mean it's _working_; it's
-    # imperative that celery workers are _actually_ handling AMQP messages on
-    # their appropriate queues for awx to function.  Unfortunately, we've been
-    # plagued by a variety of bugs in celery that cause it to hang and become
-    # an unresponsive zombie, such as:
-    #
-    # https://github.com/celery/celery/issues/4185
-    # https://github.com/celery/celery/issues/4457
-    #
-    # The goal of this code is periodically send a broadcast AMQP message to
-    # the celery process on the local host via celery.app.control.ping;
-    # If that _fails_, we attempt to determine the pid of the celery process
-    # and send SIGHUP (which tends to resolve these sorts of issues for us).
-    #
-
-    INTERVAL = 60
-
-    def _log(self, msg):
-        sys.stderr.write(datetime.datetime.utcnow().isoformat())
-        sys.stderr.write(' ')
-        sys.stderr.write(msg)
-        sys.stderr.write('\n')
-
-    def handle(self, **options):
-        app = Celery('awx')
-        app.config_from_object('django.conf:settings')
-        while True:
-            try:
-                pongs = app.control.ping(['celery@{}'.format(settings.CLUSTER_HOST_ID)], timeout=30)
-            except Exception:
-                pongs = []
-            if not pongs:
-                self._log('celery is not responsive to ping over local AMQP')
-                pid = self.getpid()
-                if pid:
-                    self._log('sending SIGHUP to {}'.format(pid))
-                    os.kill(pid, signal.SIGHUP)
-            time.sleep(self.INTERVAL)
-
-    def getpid(self):
-        cmd = 'supervisorctl pid tower-processes:awx-celeryd'
-        if os.path.exists('/supervisor_task.conf'):
-            cmd = 'supervisorctl -c /supervisor_task.conf pid tower-processes:celery'
-        try:
-            return int(subprocess.check_output(cmd, shell=True))
-        except Exception:
-            self._log('could not detect celery pid')

awx/main/middleware.py

@@ -1,25 +1,20 @@
 # Copyright (c) 2015 Ansible, Inc.
 # All Rights Reserved.
 
-import base64
-import json
+import uuid
 import logging
 import threading
-import uuid
 import six
 import time
 import cProfile
 import pstats
 import os
-import re
 
 from django.conf import settings
 from django.contrib.auth.models import User
-from django.core.exceptions import ObjectDoesNotExist
 from django.db.models.signals import post_save
 from django.db.migrations.executor import MigrationExecutor
 from django.db import IntegrityError, connection
-from django.http import HttpResponse
 from django.utils.functional import curry
 from django.shortcuts import get_object_or_404, redirect
 from django.apps import apps
@@ -209,59 +204,6 @@ class URLModificationMiddleware(object):
             request.path_info = new_path
 
 
-class DeprecatedAuthTokenMiddleware(object):
-    """
-    Used to emulate support for the old Auth Token endpoint to ease the
-    transition to OAuth2.0.  Specifically, this middleware:
-
-    1.  Intercepts POST requests to `/api/v2/authtoken/` (which now no longer
-        _actually_ exists in our urls.py)
-    2.  Rewrites `request.path` to `/api/v2/users/N/personal_tokens/`
-    3.  Detects the username and password in the request body (either in JSON,
-        or form-encoded variables) and builds an appropriate HTTP_AUTHORIZATION
-        Basic header
-    """
-
-    def process_request(self, request):
-        if re.match('^/api/v[12]/authtoken/?$', request.path):
-            if request.method != 'POST':
-                return HttpResponse('HTTP {} is not allowed.'.format(request.method), status=405)
-            try:
-                payload = json.loads(request.body)
-            except (ValueError, TypeError):
-                payload = request.POST
-            if 'username' not in payload or 'password' not in payload:
-                return HttpResponse('Unable to login with provided credentials.', status=401)
-            username = payload['username']
-            password = payload['password']
-            try:
-                pk = User.objects.get(username=username).pk
-            except ObjectDoesNotExist:
-                return HttpResponse('Unable to login with provided credentials.', status=401)
-            new_path = reverse('api:user_personal_token_list', kwargs={
-                'pk': pk,
-                'version': 'v2'
-            })
-            request._body = ''
-            request.META['CONTENT_TYPE'] = 'application/json'
-            request.path = request.path_info = new_path
-            auth = ' '.join([
-                'Basic',
-                base64.b64encode(
-                    six.text_type('{}:{}').format(username, password)
-                )
-            ])
-            request.environ['HTTP_AUTHORIZATION'] = auth
-            logger.warn(
-                'The Auth Token API (/api/v2/authtoken/) is deprecated and will '
-                'be replaced with OAuth2.0 in the next version of Ansible Tower '
-                '(see /api/o/ for more details).'
-            )
-        elif request.environ.get('HTTP_AUTHORIZATION', '').startswith('Token '):
-            token = request.environ['HTTP_AUTHORIZATION'].split(' ', 1)[-1].strip()
-            request.environ['HTTP_AUTHORIZATION'] = six.text_type('Bearer {}').format(token)
-
-
 class MigrationRanCheckMiddleware(object):
 
     def process_request(self, request):

awx/main/migrations/0050_v340_drop_celery_tables.py

@@ -0,0 +1,21 @@
+# -*- coding: utf-8 -*-
+from __future__ import unicode_literals
+
+from django.db import migrations
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('main', '0049_v330_validate_instance_capacity_adjustment'),
+    ]
+
+    operations = [
+        migrations.RunSQL([
+            ("DROP TABLE IF EXISTS {} CASCADE;".format(table))
+        ])
+        for table in ('celery_taskmeta', 'celery_tasksetmeta', 'djcelery_crontabschedule',
+                      'djcelery_intervalschedule', 'djcelery_periodictask',
+                      'djcelery_periodictasks', 'djcelery_taskstate', 'djcelery_workerstate',
+                      'djkombu_message', 'djkombu_queue')
+    ]

awx/main/migrations/0051_v340_job_slicing.py

@@ -0,0 +1,47 @@
+# -*- coding: utf-8 -*-
+# Generated by Django 1.11.11 on 2018-10-15 16:21
+from __future__ import unicode_literals
+
+import awx.main.utils.polymorphic
+from django.db import migrations, models
+import django.db.models.deletion
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('main', '0050_v340_drop_celery_tables'),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name='job',
+            name='job_slice_count',
+            field=models.PositiveIntegerField(blank=True, default=1, help_text='If ran as part of sliced jobs, the total number of slices. If 1, job is not part of a sliced job.'),
+        ),
+        migrations.AddField(
+            model_name='job',
+            name='job_slice_number',
+            field=models.PositiveIntegerField(blank=True, default=0, help_text='If part of a sliced job, the ID of the inventory slice operated on. If not part of sliced job, parameter is not used.'),
+        ),
+        migrations.AddField(
+            model_name='jobtemplate',
+            name='job_slice_count',
+            field=models.PositiveIntegerField(blank=True, default=1, help_text='The number of jobs to slice into at runtime. Will cause the Job Template to launch a workflow if value is greater than 1.'),
+        ),
+        migrations.AddField(
+            model_name='workflowjob',
+            name='is_sliced_job',
+            field=models.BooleanField(default=False),
+        ),
+        migrations.AddField(
+            model_name='workflowjob',
+            name='job_template',
+            field=models.ForeignKey(blank=True, default=None, help_text='If automatically created for a sliced job run, the job template the workflow job was created from.', null=True, on_delete=django.db.models.deletion.SET_NULL, related_name='slice_workflow_jobs', to='main.JobTemplate'),
+        ),
+        migrations.AlterField(
+            model_name='unifiedjob',
+            name='unified_job_template',
+            field=models.ForeignKey(default=None, editable=False, null=True, on_delete=awx.main.utils.polymorphic.SET_NULL, related_name='unifiedjob_unified_jobs', to='main.UnifiedJobTemplate'),
+        ),
+    ]

awx/main/models/ad_hoc_commands.py

@@ -136,8 +136,7 @@ class AdHocCommand(UnifiedJob, JobNotificationMixin):
         else:
             return []
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
         return ''
 
     @classmethod

awx/main/models/credential/__init__.py

@@ -439,15 +439,6 @@ class CredentialType(CommonModelNameNotUnique):
 
     defaults = OrderedDict()
 
-    ENV_BLACKLIST = set((
-        'VIRTUAL_ENV', 'PATH', 'PYTHONPATH', 'PROOT_TMP_DIR', 'JOB_ID',
-        'INVENTORY_ID', 'INVENTORY_SOURCE_ID', 'INVENTORY_UPDATE_ID',
-        'AD_HOC_COMMAND_ID', 'REST_API_URL', 'REST_API_TOKEN', 'MAX_EVENT_RES',
-        'CALLBACK_QUEUE', 'CALLBACK_CONNECTION', 'CACHE',
-        'JOB_CALLBACK_DEBUG', 'INVENTORY_HOSTVARS', 'FACT_QUEUE',
-        'AWX_HOST', 'PROJECT_REVISION'
-    ))
-
     class Meta:
         app_label = 'main'
         ordering = ('kind', 'name')
@@ -648,8 +639,14 @@ class CredentialType(CommonModelNameNotUnique):
                 file_label = file_label.split('.')[1]
                 setattr(tower_namespace.filename, file_label, path)
 
+        injector_field = self._meta.get_field('injectors')
         for env_var, tmpl in self.injectors.get('env', {}).items():
-            if env_var.startswith('ANSIBLE_') or env_var in self.ENV_BLACKLIST:
+            try:
+                injector_field.validate_env_var_allowed(env_var)
+            except ValidationError as e:
+                logger.error(six.text_type(
+                    'Ignoring prohibited env var {}, reason: {}'
+                ).format(env_var, e))
                 continue
             env[env_var] = Template(tmpl).render(**namespace)
             safe_env[env_var] = Template(tmpl).render(**safe_namespace)

awx/main/models/ha.py

@@ -32,7 +32,7 @@ __all__ = ('Instance', 'InstanceGroup', 'JobOrigin', 'TowerScheduleState',)
 
 
 def validate_queuename(v):
-    # celery and kombu don't play nice with unicode in queue names
+    # kombu doesn't play nice with unicode in queue names
     if v:
         try:
             '{}'.format(v.decode('utf-8'))

awx/main/models/inventory.py

@@ -9,7 +9,6 @@ import copy
 from urlparse import urljoin
 import os.path
 import six
-from distutils.version import LooseVersion
 
 # Django
 from django.conf import settings
@@ -20,6 +19,9 @@ from django.core.exceptions import ValidationError
 from django.utils.timezone import now
 from django.db.models import Q
 
+# REST Framework
+from rest_framework.exceptions import ParseError
+
 # AWX
 from awx.api.versioning import reverse
 from awx.main.constants import CLOUD_PROVIDERS
@@ -42,7 +44,7 @@ from awx.main.models.notifications import (
     NotificationTemplate,
     JobNotificationMixin,
 )
-from awx.main.utils import _inventory_updates, get_ansible_version, region_sorting
+from awx.main.utils import _inventory_updates, region_sorting
 
 
 __all__ = ['Inventory', 'Host', 'Group', 'InventorySource', 'InventoryUpdate',
@@ -218,67 +220,87 @@ class Inventory(CommonModelNameNotUnique, ResourceMixin, RelatedJobsMixin):
             group_children.add(from_group_id)
         return group_children_map
 
-    def get_script_data(self, hostvars=False, towervars=False, show_all=False):
-        if show_all:
-            hosts_q = dict()
-        else:
-            hosts_q = dict(enabled=True)
+    @staticmethod
+    def parse_slice_params(slice_str):
+        m = re.match(r"slice(?P<number>\d+)of(?P<step>\d+)", slice_str)
+        if not m:
+            raise ParseError(_('Could not parse subset as slice specification.'))
+        number = int(m.group('number'))
+        step = int(m.group('step'))
+        if number > step:
+            raise ParseError(_('Slice number must be less than total number of slices.'))
+        elif number < 1:
+            raise ParseError(_('Slice number must be 1 or higher.'))
+        return (number, step)
+
+    def get_script_data(self, hostvars=False, towervars=False, show_all=False, slice_number=1, slice_count=1):
+        hosts_kw = dict()
+        if not show_all:
+            hosts_kw['enabled'] = True
+        fetch_fields = ['name', 'id', 'variables']
+        if towervars:
+            fetch_fields.append('enabled')
+        hosts = self.hosts.filter(**hosts_kw).order_by('name').only(*fetch_fields)
+        if slice_count > 1:
+            offset = slice_number - 1
+            hosts = hosts[offset::slice_count]
+
         data = dict()
+        all_group = data.setdefault('all', dict())
+        all_hostnames = set(host.name for host in hosts)
 
         if self.variables_dict:
-            all_group = data.setdefault('all', dict())
             all_group['vars'] = self.variables_dict
+
         if self.kind == 'smart':
-            if len(self.hosts.all()) == 0:
-                return {}
-            else:
-                all_group = data.setdefault('all', dict())
-                smart_hosts_qs = self.hosts.filter(**hosts_q).all()
-                smart_hosts = list(smart_hosts_qs.values_list('name', flat=True))
-                all_group['hosts'] = smart_hosts
+            all_group['hosts'] = [host.name for host in hosts]
         else:
-            # Add hosts without a group to the all group.
-            groupless_hosts_qs = self.hosts.filter(groups__isnull=True, **hosts_q)
-            groupless_hosts = list(groupless_hosts_qs.values_list('name', flat=True))
-            if groupless_hosts:
-                all_group = data.setdefault('all', dict())
-                all_group['hosts'] = groupless_hosts
+            # Keep track of hosts that are members of a group
+            grouped_hosts = set([])
 
             # Build in-memory mapping of groups and their hosts.
-            group_hosts_kw = dict(group__inventory_id=self.id, host__inventory_id=self.id)
-            if 'enabled' in hosts_q:
-                group_hosts_kw['host__enabled'] = hosts_q['enabled']
-            group_hosts_qs = Group.hosts.through.objects.filter(**group_hosts_kw)
-            group_hosts_qs = group_hosts_qs.values_list('group_id', 'host_id', 'host__name')
+            group_hosts_qs = Group.hosts.through.objects.filter(
+                group__inventory_id=self.id,
+                host__inventory_id=self.id
+            ).values_list('group_id', 'host_id', 'host__name')
             group_hosts_map = {}
             for group_id, host_id, host_name in group_hosts_qs:
+                if host_name not in all_hostnames:
+                    continue  # host might not be in current shard
                 group_hostnames = group_hosts_map.setdefault(group_id, [])
                 group_hostnames.append(host_name)
+                grouped_hosts.add(host_name)
 
             # Build in-memory mapping of groups and their children.
             group_parents_qs = Group.parents.through.objects.filter(
                 from_group__inventory_id=self.id,
                 to_group__inventory_id=self.id,
-            )
-            group_parents_qs = group_parents_qs.values_list('from_group_id', 'from_group__name',
-                                                            'to_group_id')
+            ).values_list('from_group_id', 'from_group__name', 'to_group_id')
             group_children_map = {}
             for from_group_id, from_group_name, to_group_id in group_parents_qs:
                 group_children = group_children_map.setdefault(to_group_id, [])
                 group_children.append(from_group_name)
 
             # Now use in-memory maps to build up group info.
-            for group in self.groups.all():
+            for group in self.groups.only('name', 'id', 'variables'):
                 group_info = dict()
                 group_info['hosts'] = group_hosts_map.get(group.id, [])
                 group_info['children'] = group_children_map.get(group.id, [])
                 group_info['vars'] = group.variables_dict
                 data[group.name] = group_info
 
+            # Add ungrouped hosts to all group
+            all_group['hosts'] = [host.name for host in hosts if host.name not in grouped_hosts]
+
+        # Remove any empty groups
+        for group_name in list(data.keys()):
+            if not data.get(group_name, {}).get('hosts', []):
+                data.pop(group_name)
+
         if hostvars:
             data.setdefault('_meta', dict())
             data['_meta'].setdefault('hostvars', dict())
-            for host in self.hosts.filter(**hosts_q):
+            for host in hosts:
                 data['_meta']['hostvars'][host.name] = host.variables_dict
                 if towervars:
                     tower_dict = dict(remote_tower_enabled=str(host.enabled).lower(),
@@ -1578,12 +1600,6 @@ class InventorySource(UnifiedJobTemplate, InventorySourceOptions, RelatedJobsMix
                                     "Instead, configure the corresponding source project to update on launch."))
         return self.update_on_launch
 
-    def clean_overwrite_vars(self):  # TODO: remove when Ansible 2.4 becomes unsupported, obviously
-        if self.source == 'scm' and not self.overwrite_vars:
-            if get_ansible_version() < LooseVersion('2.5'):
-                raise ValidationError(_("SCM type sources must set `overwrite_vars` to `true` until Ansible 2.5."))
-        return self.overwrite_vars
-
     def clean_source_path(self):
         if self.source != 'scm' and self.source_path:
             raise ValidationError(_("Cannot set source_path if not SCM type."))
@@ -1631,8 +1647,7 @@ class InventoryUpdate(UnifiedJob, InventorySourceOptions, JobNotificationMixin,
         null=True
     )
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
         return 'inventory_source'
 
     @classmethod

awx/main/models/jobs.py

@@ -277,6 +277,13 @@ class JobTemplate(UnifiedJobTemplate, JobOptions, SurveyJobTemplateMixin, Resour
         default=False,
         allows_field='credentials'
     )
+    job_slice_count = models.PositiveIntegerField(
+        blank=True,
+        default=1,
+        help_text=_("The number of jobs to slice into at runtime. "
+                    "Will cause the Job Template to launch a workflow if value is greater than 1."),
+    )
+
     admin_role = ImplicitRoleField(
         parent_role=['project.organization.job_template_admin_role', 'inventory.organization.job_template_admin_role']
     )
@@ -295,7 +302,8 @@ class JobTemplate(UnifiedJobTemplate, JobOptions, SurveyJobTemplateMixin, Resour
     @classmethod
     def _get_unified_job_field_names(cls):
         return set(f.name for f in JobOptions._meta.fields) | set(
-            ['name', 'description', 'schedule', 'survey_passwords', 'labels', 'credentials']
+            ['name', 'description', 'schedule', 'survey_passwords', 'labels', 'credentials',
+             'job_slice_number', 'job_slice_count']
         )
 
     @property
@@ -320,6 +328,31 @@ class JobTemplate(UnifiedJobTemplate, JobOptions, SurveyJobTemplateMixin, Resour
         '''
         return self.create_unified_job(**kwargs)
 
+    def create_unified_job(self, **kwargs):
+        prevent_slicing = kwargs.pop('_prevent_slicing', False)
+        slice_event = bool(self.job_slice_count > 1 and (not prevent_slicing))
+        if slice_event:
+            # A Slice Job Template will generate a WorkflowJob rather than a Job
+            from awx.main.models.workflow import WorkflowJobTemplate, WorkflowJobNode
+            kwargs['_unified_job_class'] = WorkflowJobTemplate._get_unified_job_class()
+            kwargs['_parent_field_name'] = "job_template"
+            kwargs.setdefault('_eager_fields', {})
+            kwargs['_eager_fields']['is_sliced_job'] = True
+        job = super(JobTemplate, self).create_unified_job(**kwargs)
+        if slice_event:
+            try:
+                wj_config = job.launch_config
+            except JobLaunchConfig.DoesNotExist:
+                wj_config = JobLaunchConfig()
+            actual_inventory = wj_config.inventory if wj_config.inventory else self.inventory
+            for idx in xrange(min(self.job_slice_count,
+                                  actual_inventory.hosts.count())):
+                create_kwargs = dict(workflow_job=job,
+                                     unified_job_template=self,
+                                     ancestor_artifacts=dict(job_slice=idx + 1))
+                WorkflowJobNode.objects.create(**create_kwargs)
+        return job
+
     def get_absolute_url(self, request=None):
         return reverse('api:job_template_detail', kwargs={'pk': self.pk}, request=request)
 
@@ -452,7 +485,7 @@ class JobTemplate(UnifiedJobTemplate, JobOptions, SurveyJobTemplateMixin, Resour
     RelatedJobsMixin
     '''
     def _get_related_jobs(self):
-        return Job.objects.filter(job_template=self)
+        return UnifiedJob.objects.filter(unified_job_template=self)
 
 
 class Job(UnifiedJob, JobOptions, SurveyJobMixin, JobNotificationMixin, TaskManagerJobMixin):
@@ -501,10 +534,21 @@ class Job(UnifiedJob, JobOptions, SurveyJobMixin, JobNotificationMixin, TaskMana
         on_delete=models.SET_NULL,
         help_text=_('The SCM Refresh task used to make sure the playbooks were available for the job run'),
     )
+    job_slice_number = models.PositiveIntegerField(
+        blank=True,
+        default=0,
+        help_text=_("If part of a sliced job, the ID of the inventory slice operated on. "
+                    "If not part of sliced job, parameter is not used."),
+    )
+    job_slice_count = models.PositiveIntegerField(
+        blank=True,
+        default=1,
+        help_text=_("If ran as part of sliced jobs, the total number of slices. "
+                    "If 1, job is not part of a sliced job."),
+    )
 
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
         return 'job_template'
 
     @classmethod
@@ -545,6 +589,15 @@ class Job(UnifiedJob, JobOptions, SurveyJobMixin, JobNotificationMixin, TaskMana
     def event_class(self):
         return JobEvent
 
+    def copy_unified_job(self, **new_prompts):
+        # Needed for job slice relaunch consistency, do no re-spawn workflow job
+        # target same slice as original job
+        new_prompts['_prevent_slicing'] = True
+        new_prompts.setdefault('_eager_fields', {})
+        new_prompts['_eager_fields']['job_slice_number'] = self.job_slice_number
+        new_prompts['_eager_fields']['job_slice_count'] = self.job_slice_count
+        return super(Job, self).copy_unified_job(**new_prompts)
+
     @property
     def ask_diff_mode_on_launch(self):
         if self.job_template is not None:
@@ -638,6 +691,9 @@ class Job(UnifiedJob, JobOptions, SurveyJobMixin, JobNotificationMixin, TaskMana
             count_hosts = 2
         else:
             count_hosts = Host.objects.filter(inventory__jobs__pk=self.pk).count()
+            if self.job_slice_count > 1:
+                # Integer division intentional
+                count_hosts = (count_hosts + self.job_slice_count - self.job_slice_number) / self.job_slice_count
         return min(count_hosts, 5 if self.forks == 0 else self.forks) + 1
 
     @property

awx/main/models/projects.py

@@ -466,6 +466,14 @@ class Project(UnifiedJobTemplate, ProjectOptions, ResourceMixin, CustomVirtualEn
             models.Q(ProjectUpdate___project=self)
         )
 
+    def delete(self, *args, **kwargs):
+        path_to_delete = self.get_project_path(check_if_exists=False)
+        r = super(Project, self).delete(*args, **kwargs)
+        if self.scm_type and path_to_delete:  # non-manual, concrete path
+            from awx.main.tasks import delete_project_files
+            delete_project_files.delay(path_to_delete)
+        return r
+
 
 class ProjectUpdate(UnifiedJob, ProjectOptions, JobNotificationMixin, TaskManagerProjectUpdateMixin):
     '''
@@ -488,8 +496,7 @@ class ProjectUpdate(UnifiedJob, ProjectOptions, JobNotificationMixin, TaskManage
         default='check',
     )
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
         return 'project'
 
     @classmethod

awx/main/models/schedules.py

@@ -153,7 +153,7 @@ class Schedule(CommonModel, LaunchTimeConfig):
         if 'until=' in rrule.lower():
             # if DTSTART;TZID= is used, coerce "naive" UNTIL values
             # to the proper UTC date
-            match_until = re.match(".*?(?P<until>UNTIL\=[0-9]+T[0-9]+)(?P<utcflag>Z?)", rrule)
+            match_until = re.match(r".*?(?P<until>UNTIL\=[0-9]+T[0-9]+)(?P<utcflag>Z?)", rrule)
             if not len(match_until.group('utcflag')):
                 # rrule = DTSTART;TZID=America/New_York:20200601T120000 RRULE:...;UNTIL=20200601T170000
 

awx/main/models/unified_jobs.py

@@ -7,9 +7,11 @@ import json
 import logging
 import os
 import re
+import socket
 import subprocess
 import tempfile
 from collections import OrderedDict
+import six
 
 # Django
 from django.conf import settings
@@ -27,11 +29,9 @@ from rest_framework.exceptions import ParseError
 # Django-Polymorphic
 from polymorphic.models import PolymorphicModel
 
-# Django-Celery
-from djcelery.models import TaskMeta
-
 # AWX
 from awx.main.models.base import * # noqa
+from awx.main.dispatch.control import Control as ControlDispatcher
 from awx.main.models.mixins import ResourceMixin, TaskManagerUnifiedJobMixin
 from awx.main.utils import (
     encrypt_dict, decrypt_field, _inventory_updates,
@@ -309,13 +309,6 @@ class UnifiedJobTemplate(PolymorphicModel, CommonModelNameNotUnique, Notificatio
         '''
         raise NotImplementedError # Implement in subclass.
 
-    @classmethod
-    def _get_unified_job_field_names(cls):
-        '''
-        Return field names that should be copied from template to new job.
-        '''
-        raise NotImplementedError # Implement in subclass.
-
     @property
     def notification_templates(self):
         '''
@@ -338,19 +331,33 @@ class UnifiedJobTemplate(PolymorphicModel, CommonModelNameNotUnique, Notificatio
 
         unified_job_class = self._get_unified_job_class()
         fields = self._get_unified_job_field_names()
-        unallowed_fields = set(kwargs.keys()) - set(fields)
-        if unallowed_fields:
-            logger.warn('Fields {} are not allowed as overrides.'.format(unallowed_fields))
-            map(kwargs.pop, unallowed_fields)
+        parent_field_name = None
+        if "_unified_job_class" in kwargs:
+            # Special case where spawned job is different type than usual
+            # Only used for slice jobs
+            unified_job_class = kwargs.pop("_unified_job_class")
+            fields = unified_job_class._get_unified_job_field_names() & fields
+            parent_field_name = kwargs.pop('_parent_field_name')
 
-        unified_job = copy_model_by_class(self, unified_job_class, fields, kwargs)
+        unallowed_fields = set(kwargs.keys()) - set(fields)
+        validated_kwargs = kwargs.copy()
+        if unallowed_fields:
+            if parent_field_name is None:
+                logger.warn(six.text_type('Fields {} are not allowed as overrides to spawn {} from {}.').format(
+                    six.text_type(', ').join(unallowed_fields), unified_job, self
+                ))
+            map(validated_kwargs.pop, unallowed_fields)
+
+        unified_job = copy_model_by_class(self, unified_job_class, fields, validated_kwargs)
 
         if eager_fields:
             for fd, val in eager_fields.items():
                 setattr(unified_job, fd, val)
 
-        # Set the unified job template back-link on the job
-        parent_field_name = unified_job_class._get_parent_field_name()
+        # NOTE: slice workflow jobs _get_parent_field_name method
+        # is not correct until this is set
+        if not parent_field_name:
+            parent_field_name = unified_job._get_parent_field_name()
         setattr(unified_job, parent_field_name, self)
 
         # For JobTemplate-based jobs with surveys, add passwords to list for perma-redaction
@@ -364,24 +371,25 @@ class UnifiedJobTemplate(PolymorphicModel, CommonModelNameNotUnique, Notificatio
         unified_job.save()
 
         # Labels and credentials copied here
-        if kwargs.get('credentials'):
+        if validated_kwargs.get('credentials'):
             Credential = UnifiedJob._meta.get_field('credentials').related_model
             cred_dict = Credential.unique_dict(self.credentials.all())
-            prompted_dict = Credential.unique_dict(kwargs['credentials'])
+            prompted_dict = Credential.unique_dict(validated_kwargs['credentials'])
             # combine prompted credentials with JT
             cred_dict.update(prompted_dict)
-            kwargs['credentials'] = [cred for cred in cred_dict.values()]
+            validated_kwargs['credentials'] = [cred for cred in cred_dict.values()]
+            kwargs['credentials'] = validated_kwargs['credentials']
 
         from awx.main.signals import disable_activity_stream
         with disable_activity_stream():
-            copy_m2m_relationships(self, unified_job, fields, kwargs=kwargs)
+            copy_m2m_relationships(self, unified_job, fields, kwargs=validated_kwargs)
 
-        if 'extra_vars' in kwargs:
-            unified_job.handle_extra_data(kwargs['extra_vars'])
+        if 'extra_vars' in validated_kwargs:
+            unified_job.handle_extra_data(validated_kwargs['extra_vars'])
 
         if not getattr(self, '_deprecated_credential_launch', False):
             # Create record of provided prompts for relaunch and rescheduling
-            unified_job.create_config_from_prompts(kwargs)
+            unified_job.create_config_from_prompts(kwargs, parent=self)
 
         return unified_job
 
@@ -546,7 +554,7 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
         default=None,
         editable=False,
         related_name='%(class)s_unified_jobs',
-        on_delete=models.SET_NULL,
+        on_delete=polymorphic.SET_NULL,
     )
     launch_type = models.CharField(
         max_length=20,
@@ -694,8 +702,7 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
     def supports_isolation(cls):
         return False
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
         return 'unified_job_template' # Override in subclasses.
 
     @classmethod
@@ -828,7 +835,7 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
         '''
         unified_job_class = self.__class__
         unified_jt_class = self._get_unified_job_template_class()
-        parent_field_name = unified_job_class._get_parent_field_name()
+        parent_field_name = self._get_parent_field_name()
         fields = unified_jt_class._get_unified_job_field_names() | set([parent_field_name])
 
         create_data = {}
@@ -847,10 +854,10 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
                 setattr(unified_job, fd, val)
             unified_job.save()
 
-        # Labels copied here
-        from awx.main.signals import disable_activity_stream
-        with disable_activity_stream():
-            copy_m2m_relationships(self, unified_job, fields)
+            # Labels copied here
+            from awx.main.signals import disable_activity_stream
+            with disable_activity_stream():
+                copy_m2m_relationships(self, unified_job, fields)
 
         return unified_job
 
@@ -866,16 +873,18 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
         except JobLaunchConfig.DoesNotExist:
             return None
 
-    def create_config_from_prompts(self, kwargs):
+    def create_config_from_prompts(self, kwargs, parent=None):
         '''
         Create a launch configuration entry for this job, given prompts
         returns None if it can not be created
         '''
-        if self.unified_job_template is None:
-            return None
         JobLaunchConfig = self._meta.get_field('launch_config').related_model
         config = JobLaunchConfig(job=self)
-        valid_fields = self.unified_job_template.get_ask_mapping().keys()
+        if parent is None:
+            parent = getattr(self, self._get_parent_field_name())
+        if parent is None:
+            return
+        valid_fields = parent.get_ask_mapping().keys()
         # Special cases allowed for workflows
         if hasattr(self, 'extra_vars'):
             valid_fields.extend(['survey_passwords', 'extra_vars'])
@@ -892,8 +901,9 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
             setattr(config, key, value)
         config.save()
 
-        job_creds = (set(kwargs.get('credentials', [])) -
-                     set(self.unified_job_template.credentials.all()))
+        job_creds = set(kwargs.get('credentials', []))
+        if 'credentials' in [field.name for field in parent._meta.get_fields()]:
+            job_creds = job_creds - set(parent.credentials.all())
         if job_creds:
             config.credentials.add(*job_creds)
         return config
@@ -1112,14 +1122,6 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
                 pass
         return None
 
-    @property
-    def celery_task(self):
-        try:
-            if self.celery_task_id:
-                return TaskMeta.objects.get(task_id=self.celery_task_id)
-        except TaskMeta.DoesNotExist:
-            pass
-
     def get_passwords_needed_to_start(self):
         return []
 
@@ -1224,29 +1226,6 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
 
         return (True, opts)
 
-    def start_celery_task(self, opts, error_callback, success_callback, queue):
-        kwargs = {
-            'link_error': error_callback,
-            'link': success_callback,
-            'queue': None,
-            'task_id': None,
-        }
-        if not self.celery_task_id:
-            raise RuntimeError("Expected celery_task_id to be set on model.")
-        kwargs['task_id'] = self.celery_task_id
-        task_class = self._get_task_class()
-        kwargs['queue'] = queue
-        task_class().apply_async([self.pk], opts, **kwargs)
-
-    def start(self, error_callback, success_callback, **kwargs):
-        '''
-        Start the task running via Celery.
-        '''
-        (res, opts) = self.pre_start(**kwargs)
-        if res:
-            self.start_celery_task(opts, error_callback, success_callback)
-        return res
-
     def signal_start(self, **kwargs):
         """Notify the task runner system to begin work on this task."""
 
@@ -1282,46 +1261,35 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
         # Done!
         return True
 
+
+    @property
+    def actually_running(self):
+        # returns True if the job is running in the appropriate dispatcher process
+        running = False
+        if all([
+            self.status == 'running',
+            self.celery_task_id,
+            self.execution_node
+        ]):
+            # If the job is marked as running, but the dispatcher
+            # doesn't know about it (or the dispatcher doesn't reply),
+            # then cancel the job
+            timeout = 5
+            try:
+                running = self.celery_task_id in ControlDispatcher(
+                    'dispatcher', self.execution_node
+                ).running(timeout=timeout)
+            except socket.timeout:
+                logger.error(six.text_type(
+                    'could not reach dispatcher on {} within {}s'
+                ).format(self.execution_node, timeout))
+                running = False
+        return running
+
     @property
     def can_cancel(self):
         return bool(self.status in CAN_CANCEL)
 
-    def _force_cancel(self):
-        # Update the status to 'canceled' if we can detect that the job
-        # really isn't running (i.e. celery has crashed or forcefully
-        # killed the worker).
-        task_statuses = ('STARTED', 'SUCCESS', 'FAILED', 'RETRY', 'REVOKED')
-        try:
-            taskmeta = self.celery_task
-            if not taskmeta or taskmeta.status not in task_statuses:
-                return
-            from celery import current_app
-            i = current_app.control.inspect()
-            for v in (i.active() or {}).values():
-                if taskmeta.task_id in [x['id'] for x in v]:
-                    return
-            for v in (i.reserved() or {}).values():
-                if taskmeta.task_id in [x['id'] for x in v]:
-                    return
-            for v in (i.revoked() or {}).values():
-                if taskmeta.task_id in [x['id'] for x in v]:
-                    return
-            for v in (i.scheduled() or {}).values():
-                if taskmeta.task_id in [x['id'] for x in v]:
-                    return
-            instance = self.__class__.objects.get(pk=self.pk)
-            if instance.can_cancel:
-                instance.status = 'canceled'
-                update_fields = ['status']
-                if not instance.job_explanation:
-                    instance.job_explanation = 'Forced cancel'
-                    update_fields.append('job_explanation')
-                instance.save(update_fields=update_fields)
-                self.websocket_emit_status("canceled")
-        except Exception: # FIXME: Log this exception!
-            if settings.DEBUG:
-                raise
-
     def _build_job_explanation(self):
         if not self.job_explanation:
             return 'Previous Task Canceled: {"job_type": "%s", "job_name": "%s", "job_id": "%s"}' % \
@@ -1340,13 +1308,14 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
                 if self.status in ('pending', 'waiting', 'new'):
                     self.status = 'canceled'
                     cancel_fields.append('status')
+                if self.status == 'running' and not self.actually_running:
+                    self.status = 'canceled'
+                    cancel_fields.append('status')
                 if job_explanation is not None:
                     self.job_explanation = job_explanation
                     cancel_fields.append('job_explanation')
                 self.save(update_fields=cancel_fields)
                 self.websocket_emit_status("canceled")
-            if settings.BROKER_URL.startswith('amqp://'):
-                self._force_cancel()
         return self.cancel_flag
 
     @property
@@ -1402,7 +1371,7 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
                 r['{}_user_last_name'.format(name)] = created_by.last_name
         return r
 
-    def get_celery_queue_name(self):
+    def get_queue_name(self):
         return self.controller_node or self.execution_node or settings.CELERY_DEFAULT_QUEUE
 
     def is_isolated(self):

awx/main/models/workflow.py

@@ -9,6 +9,7 @@ import logging
 from django.db import models
 from django.conf import settings
 from django.utils.translation import ugettext_lazy as _
+from django.core.exceptions import ObjectDoesNotExist
 #from django import settings as tower_settings
 
 # AWX
@@ -206,13 +207,24 @@ class WorkflowJobNode(WorkflowNodeBase):
                                   workflow_pk=self.pk,
                                   error_text=errors))
             data.update(accepted_fields)  # missing fields are handled in the scheduler
+            try:
+                # config saved on the workflow job itself
+                wj_config = self.workflow_job.launch_config
+            except ObjectDoesNotExist:
+                wj_config = None
+            if wj_config:
+                accepted_fields, ignored_fields, errors = ujt_obj._accept_or_ignore_job_kwargs(**wj_config.prompts_dict())
+                accepted_fields.pop('extra_vars', None)  # merge handled with other extra_vars later
+                data.update(accepted_fields)
         # build ancestor artifacts, save them to node model for later
         aa_dict = {}
+        is_root_node = True
         for parent_node in self.get_parent_nodes():
+            is_root_node = False
             aa_dict.update(parent_node.ancestor_artifacts)
             if parent_node.job and hasattr(parent_node.job, 'artifacts'):
                 aa_dict.update(parent_node.job.artifacts)
-        if aa_dict:
+        if aa_dict and not is_root_node:
             self.ancestor_artifacts = aa_dict
             self.save(update_fields=['ancestor_artifacts'])
         # process password list
@@ -240,6 +252,12 @@ class WorkflowJobNode(WorkflowNodeBase):
             data['extra_vars'] = extra_vars
         # ensure that unified jobs created by WorkflowJobs are marked
         data['_eager_fields'] = {'launch_type': 'workflow'}
+        # Extra processing in the case that this is a slice job
+        if 'job_slice' in self.ancestor_artifacts and is_root_node:
+            data['_eager_fields']['allow_simultaneous'] = True
+            data['_eager_fields']['job_slice_number'] = self.ancestor_artifacts['job_slice']
+            data['_eager_fields']['job_slice_count'] = self.workflow_job.workflow_job_nodes.count()
+            data['_prevent_slicing'] = True
         return data
 
 
@@ -261,6 +279,12 @@ class WorkflowJobOptions(BaseModel):
     def workflow_nodes(self):
         raise NotImplementedError()
 
+    @classmethod
+    def _get_unified_job_field_names(cls):
+        return set(f.name for f in WorkflowJobOptions._meta.fields) | set(
+            ['name', 'description', 'schedule', 'survey_passwords', 'labels']
+        )
+
     def _create_workflow_nodes(self, old_node_list, user=None):
         node_links = {}
         for old_node in old_node_list:
@@ -288,7 +312,7 @@ class WorkflowJobOptions(BaseModel):
 
     def create_relaunch_workflow_job(self):
         new_workflow_job = self.copy_unified_job()
-        if self.workflow_job_template is None:
+        if self.unified_job_template_id is None:
             new_workflow_job.copy_nodes_from_original(original=self)
         return new_workflow_job
 
@@ -331,12 +355,6 @@ class WorkflowJobTemplate(UnifiedJobTemplate, WorkflowJobOptions, SurveyJobTempl
     def _get_unified_job_class(cls):
         return WorkflowJob
 
-    @classmethod
-    def _get_unified_job_field_names(cls):
-        return set(f.name for f in WorkflowJobOptions._meta.fields) | set(
-            ['name', 'description', 'schedule', 'survey_passwords', 'labels']
-        )
-
     @classmethod
     def _get_unified_jt_copy_names(cls):
         base_list = super(WorkflowJobTemplate, cls)._get_unified_jt_copy_names()
@@ -433,13 +451,28 @@ class WorkflowJob(UnifiedJob, WorkflowJobOptions, SurveyJobMixin, JobNotificatio
         default=None,
         on_delete=models.SET_NULL,
     )
+    job_template = models.ForeignKey(
+        'JobTemplate',
+        related_name='slice_workflow_jobs',
+        blank=True,
+        null=True,
+        default=None,
+        on_delete=models.SET_NULL,
+        help_text=_("If automatically created for a sliced job run, the job template "
+                    "the workflow job was created from."),
+    )
+    is_sliced_job = models.BooleanField(
+        default=False
+    )
 
     @property
     def workflow_nodes(self):
         return self.workflow_job_nodes
 
-    @classmethod
-    def _get_parent_field_name(cls):
+    def _get_parent_field_name(self):
+        if self.job_template_id:
+            # This is a workflow job which is a container for slice jobs
+            return 'job_template'
         return 'workflow_job_template'
 
     @classmethod
@@ -482,8 +515,8 @@ class WorkflowJob(UnifiedJob, WorkflowJobOptions, SurveyJobMixin, JobNotificatio
     def preferred_instance_groups(self):
         return []
 
-    '''
-    A WorkflowJob is a virtual job. It doesn't result in a celery task.
-    '''
-    def start_celery_task(self, opts, error_callback, success_callback, queue):
-        return None
+    @property
+    def actually_running(self):
+        # WorkflowJobs don't _actually_ run anything in the dispatcher, so
+        # there's no point in asking the dispatcher if it knows about this task
+        return self.status == 'running'

awx/main/scheduler/dag_workflow.py

@@ -50,6 +50,7 @@ class WorkflowDAG(SimpleDAG):
         return [n['node_object'] for n in nodes_found]
 
     def cancel_node_jobs(self):
+        cancel_finished = True
         for n in self.nodes:
             obj = n['node_object']
             job = obj.job
@@ -57,7 +58,9 @@ class WorkflowDAG(SimpleDAG):
             if not job:
                 continue
             elif job.can_cancel:
+                cancel_finished = False
                 job.cancel()
+        return cancel_finished
 
     def is_workflow_done(self):
         root_nodes = self.get_root_nodes()

awx/main/scheduler/task_manager.py

@@ -2,7 +2,7 @@
 # All Rights Reserved
 
 # Python
-from datetime import datetime, timedelta
+from datetime import timedelta
 import logging
 import uuid
 import json
@@ -11,18 +11,13 @@ import random
 from sets import Set
 
 # Django
-from django.conf import settings
-from django.core.cache import cache
-from django.db import transaction, connection, DatabaseError
+from django.db import transaction, connection
 from django.utils.translation import ugettext_lazy as _
-from django.utils.timezone import now as tz_now, utc
-from django.db.models import Q
-from django.contrib.contenttypes.models import ContentType
+from django.utils.timezone import now as tz_now
 
 # AWX
 from awx.main.models import (
     AdHocCommand,
-    Instance,
     InstanceGroup,
     InventorySource,
     InventoryUpdate,
@@ -30,21 +25,15 @@ from awx.main.models import (
     Project,
     ProjectUpdate,
     SystemJob,
-    UnifiedJob,
     WorkflowJob,
 )
 from awx.main.scheduler.dag_workflow import WorkflowDAG
 from awx.main.utils.pglock import advisory_lock
 from awx.main.utils import get_type_for_model
 from awx.main.signals import disable_activity_stream
-
 from awx.main.scheduler.dependency_graph import DependencyGraph
 from awx.main.utils import decrypt_field
 
-# Celery
-from celery import Celery
-from celery.app.control import Inspect
-
 
 logger = logging.getLogger('awx.main.scheduler')
 
@@ -85,79 +74,6 @@ class TaskManager():
                            key=lambda task: task.created)
         return all_tasks
 
-    '''
-    Tasks that are running and SHOULD have a celery task.
-    {
-        'execution_node': [j1, j2,...],
-        'execution_node': [j3],
-        ...
-    }
-    '''
-    def get_running_tasks(self):
-        execution_nodes = {}
-        waiting_jobs = []
-        now = tz_now()
-        workflow_ctype_id = ContentType.objects.get_for_model(WorkflowJob).id
-        jobs = UnifiedJob.objects.filter((Q(status='running') |
-                                         Q(status='waiting', modified__lte=now - timedelta(seconds=60))) &
-                                         ~Q(polymorphic_ctype_id=workflow_ctype_id))
-        for j in jobs:
-            if j.execution_node:
-                execution_nodes.setdefault(j.execution_node, []).append(j)
-            else:
-                waiting_jobs.append(j)
-        return (execution_nodes, waiting_jobs)
-
-    '''
-    Tasks that are currently running in celery
-
-    Transform:
-    {
-        "celery@ec2-54-204-222-62.compute-1.amazonaws.com": [],
-        "celery@ec2-54-163-144-168.compute-1.amazonaws.com": [{
-            ...
-            "id": "5238466a-f8c7-43b3-9180-5b78e9da8304",
-            ...
-        }, {
-            ...,
-        }, ...]
-    }
-
-    to:
-    {
-        "ec2-54-204-222-62.compute-1.amazonaws.com": [
-            "5238466a-f8c7-43b3-9180-5b78e9da8304",
-            "5238466a-f8c7-43b3-9180-5b78e9da8306",
-            ...
-        ]
-    }
-    '''
-    def get_active_tasks(self):
-        if not hasattr(settings, 'IGNORE_CELERY_INSPECTOR'):
-            app = Celery('awx')
-            app.config_from_object('django.conf:settings')
-            inspector = Inspect(app=app)
-            active_task_queues = inspector.active()
-        else:
-            logger.warn("Ignoring celery task inspector")
-            active_task_queues = None
-
-        queues = None
-
-        if active_task_queues is not None:
-            queues = {}
-            for queue in active_task_queues:
-                active_tasks = set()
-                map(lambda at: active_tasks.add(at['id']), active_task_queues[queue])
-
-                # celery worker name is of the form celery@myhost.com
-                queue_name = queue.split('@')
-                queue_name = queue_name[1 if len(queue_name) > 1 else 0]
-                queues[queue_name] = active_tasks
-        else:
-            return (None, None)
-
-        return (active_task_queues, queues)
 
     def get_latest_project_update_tasks(self, all_sorted_tasks):
         project_ids = Set()
@@ -187,8 +103,15 @@ class TaskManager():
 
     def spawn_workflow_graph_jobs(self, workflow_jobs):
         for workflow_job in workflow_jobs:
+            if workflow_job.cancel_flag:
+                logger.debug('Not spawning jobs for %s because it is pending cancelation.', workflow_job.log_format)
+                continue
             dag = WorkflowDAG(workflow_job)
             spawn_nodes = dag.bfs_nodes_to_run()
+            if spawn_nodes:
+                logger.info('Spawning jobs for %s', workflow_job.log_format)
+            else:
+                logger.debug('No nodes to spawn for %s', workflow_job.log_format)
             for spawn_node in spawn_nodes:
                 if spawn_node.unified_job_template is None:
                     continue
@@ -196,8 +119,13 @@ class TaskManager():
                 job = spawn_node.unified_job_template.create_unified_job(**kv)
                 spawn_node.job = job
                 spawn_node.save()
+                logger.info('Spawned %s in %s for node %s', job.log_format, workflow_job.log_format, spawn_node.pk)
                 if job._resources_sufficient_for_launch():
-                    can_start = job.signal_start()
+                    if workflow_job.start_args:
+                        start_args = json.loads(decrypt_field(workflow_job, 'start_args'))
+                    else:
+                        start_args = {}
+                    can_start = job.signal_start(**start_args)
                     if not can_start:
                         job.job_explanation = _("Job spawned from workflow could not start because it "
                                                 "was not in the right state or required manual credentials")
@@ -213,23 +141,30 @@ class TaskManager():
                 # TODO: should we emit a status on the socket here similar to tasks.py awx_periodic_scheduler() ?
                 #emit_websocket_notification('/socket.io/jobs', '', dict(id=))
 
-    # See comment in tasks.py::RunWorkflowJob::run()
     def process_finished_workflow_jobs(self, workflow_jobs):
         result = []
         for workflow_job in workflow_jobs:
             dag = WorkflowDAG(workflow_job)
             if workflow_job.cancel_flag:
-                workflow_job.status = 'canceled'
-                workflow_job.save()
-                dag.cancel_node_jobs()
-                connection.on_commit(lambda: workflow_job.websocket_emit_status(workflow_job.status))
+                logger.debug('Canceling spawned jobs of %s due to cancel flag.', workflow_job.log_format)
+                cancel_finished = dag.cancel_node_jobs()
+                if cancel_finished:
+                    logger.info('Marking %s as canceled, all spawned jobs have concluded.', workflow_job.log_format)
+                    workflow_job.status = 'canceled'
+                    workflow_job.start_args = ''  # blank field to remove encrypted passwords
+                    workflow_job.save(update_fields=['status', 'start_args'])
+                    connection.on_commit(lambda: workflow_job.websocket_emit_status(workflow_job.status))
             else:
                 is_done, has_failed = dag.is_workflow_done()
                 if not is_done:
                     continue
+                logger.info('Marking %s as %s.', workflow_job.log_format, 'failed' if has_failed else 'successful')
                 result.append(workflow_job.id)
-                workflow_job.status = 'failed' if has_failed else 'successful'
-                workflow_job.save()
+                new_status = 'failed' if has_failed else 'successful'
+                logger.debug(six.text_type("Transitioning {} to {} status.").format(workflow_job.log_format, new_status))
+                workflow_job.status = new_status
+                workflow_job.start_args = ''  # blank field to remove encrypted passwords
+                workflow_job.save(update_fields=['status', 'start_args'])
                 connection.on_commit(lambda: workflow_job.websocket_emit_status(workflow_job.status))
         return result
 
@@ -256,9 +191,6 @@ class TaskManager():
                              rampart_group.name, task.log_format))
                 return
 
-        error_handler = handle_work_error.s(subtasks=[task_actual] + dependencies)
-        success_handler = handle_work_success.s(task_actual=task_actual)
-
         task.status = 'waiting'
 
         (start_status, opts) = task.pre_start()
@@ -300,11 +232,23 @@ class TaskManager():
 
         def post_commit():
             task.websocket_emit_status(task.status)
-            if task.status != 'failed':
-                task.start_celery_task(opts,
-                                       error_callback=error_handler,
-                                       success_callback=success_handler,
-                                       queue=task.get_celery_queue_name())
+            if task.status != 'failed' and type(task) is not WorkflowJob:
+                task_cls = task._get_task_class()
+                task_cls.apply_async(
+                    [task.pk],
+                    opts,
+                    queue=task.get_queue_name(),
+                    uuid=task.celery_task_id,
+                    callbacks=[{
+                        'task': handle_work_success.name,
+                        'kwargs': {'task_actual': task_actual}
+                    }],
+                    errbacks=[{
+                        'task': handle_work_error.name,
+                        'args': [task.celery_task_id],
+                        'kwargs': {'subtasks': [task_actual] + dependencies}
+                    }],
+                )
 
         connection.on_commit(post_commit)
 
@@ -483,7 +427,7 @@ class TaskManager():
                 logger.debug(six.text_type("Dependent {} couldn't be scheduled on graph, waiting for next cycle").format(task.log_format))
 
     def process_pending_tasks(self, pending_tasks):
-        running_workflow_templates = set([wf.workflow_job_template_id for wf in self.get_running_workflow_jobs()])
+        running_workflow_templates = set([wf.unified_job_template_id for wf in self.get_running_workflow_jobs()])
         for task in pending_tasks:
             self.process_dependencies(task, self.generate_dependencies(task))
             if self.is_job_blocked(task):
@@ -493,12 +437,12 @@ class TaskManager():
             found_acceptable_queue = False
             idle_instance_that_fits = None
             if isinstance(task, WorkflowJob):
-                if task.workflow_job_template_id in running_workflow_templates:
+                if task.unified_job_template_id in running_workflow_templates:
                     if not task.allow_simultaneous:
                         logger.debug(six.text_type("{} is blocked from running, workflow already running").format(task.log_format))
                         continue
                 else:
-                    running_workflow_templates.add(task.workflow_job_template_id)
+                    running_workflow_templates.add(task.unified_job_template_id)
                 self.start_task(task, None, task.get_jobs_fail_chain(), None)
                 continue
             for rampart_group in preferred_instance_groups:
@@ -529,105 +473,6 @@ class TaskManager():
             if not found_acceptable_queue:
                 logger.debug(six.text_type("{} couldn't be scheduled on graph, waiting for next cycle").format(task.log_format))
 
-    def fail_jobs_if_not_in_celery(self, node_jobs, active_tasks, celery_task_start_time,
-                                   isolated=False):
-        for task in node_jobs:
-            if (task.celery_task_id not in active_tasks and not hasattr(settings, 'IGNORE_CELERY_INSPECTOR')):
-                if isinstance(task, WorkflowJob):
-                    continue
-                if task.modified > celery_task_start_time:
-                    continue
-                new_status = 'failed'
-                if isolated:
-                    new_status = 'error'
-                task.status = new_status
-                task.start_args = ''  # blank field to remove encrypted passwords
-                if isolated:
-                    # TODO: cancel and reap artifacts of lost jobs from heartbeat
-                    task.job_explanation += ' '.join((
-                        'Task was marked as running in Tower but its ',
-                        'controller management daemon was not present in',
-                        'the job queue, so it has been marked as failed.',
-                        'Task may still be running, but contactability is unknown.'
-                    ))
-                else:
-                    task.job_explanation += ' '.join((
-                        'Task was marked as running in Tower but was not present in',
-                        'the job queue, so it has been marked as failed.',
-                    ))
-                try:
-                    task.save(update_fields=['status', 'start_args', 'job_explanation'])
-                except DatabaseError:
-                    logger.error("Task {} DB error in marking failed. Job possibly deleted.".format(task.log_format))
-                    continue
-                if hasattr(task, 'send_notification_templates'):
-                    task.send_notification_templates('failed')
-                task.websocket_emit_status(new_status)
-                logger.error("{}Task {} has no record in celery. Marking as failed".format(
-                    'Isolated ' if isolated else '', task.log_format))
-
-    def cleanup_inconsistent_celery_tasks(self):
-        '''
-        Rectify tower db <-> celery inconsistent view of jobs state
-        '''
-        last_cleanup = cache.get('last_celery_task_cleanup') or datetime.min.replace(tzinfo=utc)
-        if (tz_now() - last_cleanup).seconds < settings.AWX_INCONSISTENT_TASK_INTERVAL:
-            return
-
-        logger.debug("Failing inconsistent running jobs.")
-        celery_task_start_time = tz_now()
-        active_task_queues, active_queues = self.get_active_tasks()
-        cache.set('last_celery_task_cleanup', tz_now())
-
-        if active_queues is None:
-            logger.error('Failed to retrieve active tasks from celery')
-            return None
-
-        '''
-        Only consider failing tasks on instances for which we obtained a task
-        list from celery for.
-        '''
-        running_tasks, waiting_tasks = self.get_running_tasks()
-        all_celery_task_ids = []
-        for node, node_jobs in active_queues.iteritems():
-            all_celery_task_ids.extend(node_jobs)
-
-        self.fail_jobs_if_not_in_celery(waiting_tasks, all_celery_task_ids, celery_task_start_time)
-
-        for node, node_jobs in running_tasks.iteritems():
-            isolated = False
-            if node in active_queues:
-                active_tasks = active_queues[node]
-            else:
-                '''
-                Node task list not found in celery. We may branch into cases:
-                 - instance is unknown to tower, system is improperly configured
-                 - instance is reported as down, then fail all jobs on the node
-                 - instance is an isolated node, then check running tasks
-                   among all allowed controller nodes for management process
-                 - valid healthy instance not included in celery task list
-                   probably a netsplit case, leave it alone
-                '''
-                instance = Instance.objects.filter(hostname=node).first()
-
-                if instance is None:
-                    logger.error("Execution node Instance {} not found in database. "
-                                 "The node is currently executing jobs {}".format(
-                                     node, [j.log_format for j in node_jobs]))
-                    active_tasks = []
-                elif instance.capacity == 0:
-                    active_tasks = []
-                elif instance.rampart_groups.filter(controller__isnull=False).exists():
-                    active_tasks = all_celery_task_ids
-                    isolated = True
-                else:
-                    continue
-
-            self.fail_jobs_if_not_in_celery(
-                node_jobs, active_tasks, celery_task_start_time,
-                isolated=isolated
-            )
-
     def calculate_capacity_consumed(self, tasks):
         self.graph = InstanceGroup.objects.capacity_values(tasks=tasks, graph=self.graph)
 
@@ -673,6 +518,14 @@ class TaskManager():
             running_workflow_tasks = self.get_running_workflow_jobs()
             finished_wfjs = self.process_finished_workflow_jobs(running_workflow_tasks)
 
+            previously_running_workflow_tasks = running_workflow_tasks
+            running_workflow_tasks = []
+            for workflow_job in previously_running_workflow_tasks:
+                if workflow_job.status == 'running':
+                    running_workflow_tasks.append(workflow_job)
+                else:
+                    logger.debug('Removed %s from job spawning consideration.', workflow_job.log_format)
+
             self.spawn_workflow_graph_jobs(running_workflow_tasks)
 
             self.process_tasks(all_sorted_tasks)
@@ -687,7 +540,6 @@ class TaskManager():
                     return
                 logger.debug("Starting Scheduler")
 
-                self.cleanup_inconsistent_celery_tasks()
                 finished_wfjs = self._schedule()
 
                 # Operations whose queries rely on modifications made during the atomic scheduling session

awx/main/scheduler/tasks.py

@@ -2,30 +2,24 @@
 # Python
 import logging
 
-# Celery
-from celery import shared_task
-
 # AWX
 from awx.main.scheduler import TaskManager
+from awx.main.dispatch.publish import task
 
 logger = logging.getLogger('awx.main.scheduler')
 
-# TODO: move logic to UnifiedJob model and use bind=True feature of celery.
-# Would we need the request loop then? I think so. Even if we get the in-memory
-# updated model, the call to schedule() may get stale data.
 
-
-@shared_task()
+@task()
 def run_job_launch(job_id):
     TaskManager().schedule()
 
 
-@shared_task()
+@task()
 def run_job_complete(job_id):
     TaskManager().schedule()
 
 
-@shared_task()
+@task()
 def run_task_manager():
     logger.debug("Running Tower task manager.")
     TaskManager().schedule()

awx/main/tasks.py

@@ -13,12 +13,11 @@ import logging
 import os
 import re
 import shutil
+import six
 import stat
-import sys
 import tempfile
 import time
 import traceback
-import six
 import urlparse
 from distutils.version import LooseVersion as Version
 import yaml
@@ -28,12 +27,6 @@ try:
 except Exception:
     psutil = None
 
-# Celery
-from kombu import Queue, Exchange
-from kombu.common import Broadcast
-from celery import Task, shared_task
-from celery.signals import celeryd_init, worker_shutdown
-
 # Django
 from django.conf import settings
 from django.db import transaction, DatabaseError, IntegrityError
@@ -58,10 +51,12 @@ from awx.main.constants import ACTIVE_STATES
 from awx.main.exceptions import AwxTaskError
 from awx.main.queue import CallbackQueueDispatcher
 from awx.main.expect import run, isolated_manager
+from awx.main.dispatch.publish import task
+from awx.main.dispatch import get_local_queuename, reaper
 from awx.main.utils import (get_ansible_version, get_ssh_version, decrypt_field, update_scm_url,
                             check_proot_installed, build_proot_temp_dir, get_licenser,
                             wrap_args_with_proot, OutputEventFilter, OutputVerboseFilter, ignore_inventory_computed_fields,
-                            ignore_inventory_group_removal, get_type_for_model, extract_ansible_vars)
+                            ignore_inventory_group_removal, extract_ansible_vars)
 from awx.main.utils.safe_yaml import safe_dump, sanitize_jinja
 from awx.main.utils.reload import stop_local_services
 from awx.main.utils.pglock import advisory_lock
@@ -87,52 +82,7 @@ Try upgrading OpenSSH or providing your private key in an different format. \
 logger = logging.getLogger('awx.main.tasks')
 
 
-def log_celery_failure(self, exc, task_id, args, kwargs, einfo):
-    try:
-        if getattr(exc, 'is_awx_task_error', False):
-            # Error caused by user / tracked in job output
-            logger.warning(six.text_type("{}").format(exc))
-        elif isinstance(self, BaseTask):
-            logger.exception(six.text_type(
-                '{!s} {!s} execution encountered exception.')
-                .format(get_type_for_model(self.model), args[0]))
-        else:
-            logger.exception(six.text_type('Task {} encountered exception.').format(self.name), exc_info=exc)
-    except Exception:
-        # It's fairly critical that this code _not_ raise exceptions on logging
-        # If you configure external logging in a way that _it_ fails, there's
-        # not a lot we can do here; sys.stderr.write is a final hail mary
-        _, _, tb = sys.exc_info()
-        traceback.print_tb(tb)
-
-
-@celeryd_init.connect
-def celery_startup(conf=None, **kwargs):
-    #
-    # When celeryd starts, if the instance cannot be found in the database,
-    # automatically register it.  This is mostly useful for openshift-based
-    # deployments where:
-    #
-    # 2 Instances come online
-    # Instance B encounters a network blip, Instance A notices, and
-    # deprovisions it
-    # Instance B's connectivity is restored, celeryd starts, and it
-    # re-registers itself
-    #
-    # In traditional container-less deployments, instances don't get
-    # deprovisioned when they miss their heartbeat, so this code is mostly a
-    # no-op.
-    #
-    if kwargs['instance'].hostname != 'celery@{}'.format(settings.CLUSTER_HOST_ID):
-        error = six.text_type('celery -n {} does not match settings.CLUSTER_HOST_ID={}').format(
-            instance.hostname, settings.CLUSTER_HOST_ID
-        )
-        logger.error(error)
-        raise RuntimeError(error)
-    (changed, tower_instance) = Instance.objects.get_or_register()
-    if changed:
-        logger.info(six.text_type("Registered tower node '{}'").format(tower_instance.hostname))
-
+def dispatch_startup():
     startup_logger = logging.getLogger('awx.main.tasks')
     startup_logger.info("Syncing Schedules")
     for sch in Schedule.objects.all():
@@ -144,34 +94,44 @@ def celery_startup(conf=None, **kwargs):
         except Exception:
             logger.exception(six.text_type("Failed to rebuild schedule {}.").format(sch))
 
-    # set the queues we want to bind to dynamically at startup
-    queues = []
-    me = Instance.objects.me()
-    for q in [me.hostname] + settings.AWX_CELERY_QUEUES_STATIC:
-        q = q.encode('utf-8')
-        queues.append(Queue(q, Exchange(q), routing_key=q))
-    for q in settings.AWX_CELERY_BCAST_QUEUES_STATIC:
-        queues.append(Broadcast(q.encode('utf-8')))
-    conf.CELERY_QUEUES = list(set(queues))
-
-    # Expedite the first hearbeat run so a node comes online quickly.
-    cluster_node_heartbeat.apply([])
+    #
+    # When the dispatcher starts, if the instance cannot be found in the database,
+    # automatically register it.  This is mostly useful for openshift-based
+    # deployments where:
+    #
+    # 2 Instances come online
+    # Instance B encounters a network blip, Instance A notices, and
+    # deprovisions it
+    # Instance B's connectivity is restored, the dispatcher starts, and it
+    # re-registers itself
+    #
+    # In traditional container-less deployments, instances don't get
+    # deprovisioned when they miss their heartbeat, so this code is mostly a
+    # no-op.
+    #
+    apply_cluster_membership_policies()
+    cluster_node_heartbeat()
+    if Instance.objects.me().is_controller():
+        awx_isolated_heartbeat()
 
 
-@worker_shutdown.connect
-def inform_cluster_of_shutdown(*args, **kwargs):
+def inform_cluster_of_shutdown():
     try:
         this_inst = Instance.objects.get(hostname=settings.CLUSTER_HOST_ID)
         this_inst.capacity = 0  # No thank you to new jobs while shut down
         this_inst.save(update_fields=['capacity', 'modified'])
+        try:
+            reaper.reap(this_inst)
+        except Exception:
+            logger.exception('failed to reap jobs for {}'.format(this_inst.hostname))
         logger.warning(six.text_type('Normal shutdown signal for instance {}, '
                        'removed self from capacity pool.').format(this_inst.hostname))
     except Exception:
         logger.exception('Encountered problem with normal shutdown signal.')
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE)
-def apply_cluster_membership_policies(self):
+@task()
+def apply_cluster_membership_policies():
     started_waiting = time.time()
     with advisory_lock('cluster_policy_lock', wait=True):
         lock_time = time.time() - started_waiting
@@ -280,20 +240,36 @@ def apply_cluster_membership_policies(self):
         logger.info('Cluster policy computation finished in {} seconds'.format(time.time() - started_compute))
 
 
-@shared_task(exchange='tower_broadcast_all', bind=True)
-def handle_setting_changes(self, setting_keys):
+@task(queue='tower_broadcast_all', exchange_type='fanout')
+def handle_setting_changes(setting_keys):
     orig_len = len(setting_keys)
     for i in range(orig_len):
         for dependent_key in settings_registry.get_dependent_settings(setting_keys[i]):
             setting_keys.append(dependent_key)
-    logger.warn('Processing cache changes, task args: {0.args!r} kwargs: {0.kwargs!r}'.format(
-        self.request))
     cache_keys = set(setting_keys)
     logger.debug('cache delete_many(%r)', cache_keys)
     cache.delete_many(cache_keys)
 
 
-@shared_task(queue=settings.CELERY_DEFAULT_QUEUE)
+@task(queue='tower_broadcast_all', exchange_type='fanout')
+def delete_project_files(project_path):
+    # TODO: possibly implement some retry logic
+    lock_file = project_path + '.lock'
+    if os.path.exists(project_path):
+        try:
+            shutil.rmtree(project_path)
+            logger.info(six.text_type('Success removing project files {}').format(project_path))
+        except Exception:
+            logger.exception(six.text_type('Could not remove project directory {}').format(project_path))
+    if os.path.exists(lock_file):
+        try:
+            os.remove(lock_file)
+            logger.debug(six.text_type('Success removing {}').format(lock_file))
+        except Exception:
+            logger.exception(six.text_type('Could not remove lock file {}').format(lock_file))
+
+
+@task()
 def send_notifications(notification_list, job_id=None):
     if not isinstance(notification_list, list):
         raise TypeError("notification_list should be of type list")
@@ -322,8 +298,8 @@ def send_notifications(notification_list, job_id=None):
                 logger.exception(six.text_type('Error saving notification {} result.').format(notification.id))
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE)
-def run_administrative_checks(self):
+@task()
+def run_administrative_checks():
     logger.warn("Running administrative checks.")
     if not settings.TOWER_ADMIN_ALERTS:
         return
@@ -344,8 +320,8 @@ def run_administrative_checks(self):
                   fail_silently=True)
 
 
-@shared_task(bind=True)
-def purge_old_stdout_files(self):
+@task(queue=get_local_queuename)
+def purge_old_stdout_files():
     nowtime = time.time()
     for f in os.listdir(settings.JOBOUTPUT_ROOT):
         if os.path.getctime(os.path.join(settings.JOBOUTPUT_ROOT,f)) < nowtime - settings.LOCAL_STDOUT_EXPIRE_TIME:
@@ -353,8 +329,8 @@ def purge_old_stdout_files(self):
             logger.info(six.text_type("Removing {}").format(os.path.join(settings.JOBOUTPUT_ROOT,f)))
 
 
-@shared_task(bind=True)
-def cluster_node_heartbeat(self):
+@task(queue=get_local_queuename)
+def cluster_node_heartbeat():
     logger.debug("Cluster node heartbeat task.")
     nowtime = now()
     instance_list = list(Instance.objects.all_non_isolated())
@@ -397,9 +373,13 @@ def cluster_node_heartbeat(self):
                                     this_inst.version))
             # Shutdown signal will set the capacity to zero to ensure no Jobs get added to this instance.
             # The heartbeat task will reset the capacity to the system capacity after upgrade.
-            stop_local_services(['uwsgi', 'celery', 'beat', 'callback'], communicate=False)
+            stop_local_services(communicate=False)
             raise RuntimeError("Shutting down.")
     for other_inst in lost_instances:
+        try:
+            reaper.reap(other_inst)
+        except Exception:
+            logger.exception('failed to reap jobs for {}'.format(other_inst.hostname))
         try:
             # Capacity could already be 0 because:
             #  * It's a new node and it never had a heartbeat
@@ -424,8 +404,8 @@ def cluster_node_heartbeat(self):
                 logger.exception(six.text_type('Error marking {} as lost').format(other_inst.hostname))
 
 
-@shared_task(bind=True)
-def awx_isolated_heartbeat(self):
+@task(queue=get_local_queuename)
+def awx_isolated_heartbeat():
     local_hostname = settings.CLUSTER_HOST_ID
     logger.debug("Controlling node checking for any isolated management tasks.")
     poll_interval = settings.AWX_ISOLATED_PERIODIC_CHECK
@@ -452,8 +432,8 @@ def awx_isolated_heartbeat(self):
         isolated_manager.IsolatedManager.health_check(isolated_instance_qs, awx_application_version)
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE)
-def awx_periodic_scheduler(self):
+@task()
+def awx_periodic_scheduler():
     run_now = now()
     state = TowerScheduleState.get_solo()
     last_run = state.schedule_last_run
@@ -503,8 +483,8 @@ def awx_periodic_scheduler(self):
     state.save()
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE)
-def handle_work_success(self, result, task_actual):
+@task()
+def handle_work_success(task_actual):
     try:
         instance = UnifiedJob.get_instance_by_type(task_actual['type'], task_actual['id'])
     except ObjectDoesNotExist:
@@ -517,7 +497,7 @@ def handle_work_success(self, result, task_actual):
     run_job_complete.delay(instance.id)
 
 
-@shared_task(queue=settings.CELERY_DEFAULT_QUEUE)
+@task()
 def handle_work_error(task_id, *args, **kwargs):
     subtasks = kwargs.get('subtasks', None)
     logger.debug('Executing error task id %s, subtasks: %s' % (task_id, str(subtasks)))
@@ -558,7 +538,7 @@ def handle_work_error(task_id, *args, **kwargs):
         pass
 
 
-@shared_task(queue=settings.CELERY_DEFAULT_QUEUE)
+@task()
 def update_inventory_computed_fields(inventory_id, should_update_hosts=True):
     '''
     Signal handler and wrapper around inventory.update_computed_fields to
@@ -578,7 +558,7 @@ def update_inventory_computed_fields(inventory_id, should_update_hosts=True):
         raise
 
 
-@shared_task(queue=settings.CELERY_DEFAULT_QUEUE)
+@task()
 def update_host_smart_inventory_memberships():
     try:
         with transaction.atomic():
@@ -603,8 +583,8 @@ def update_host_smart_inventory_memberships():
         smart_inventory.update_computed_fields(update_groups=False, update_hosts=False)
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE, max_retries=5)
-def delete_inventory(self, inventory_id, user_id):
+@task()
+def delete_inventory(inventory_id, user_id, retries=5):
     # Delete inventory as user
     if user_id is None:
         user = None
@@ -629,7 +609,9 @@ def delete_inventory(self, inventory_id, user_id):
             return
         except DatabaseError:
             logger.exception('Database error deleting inventory {}, but will retry.'.format(inventory_id))
-            self.retry(countdown=10)
+            if retries > 0:
+                time.sleep(10)
+                delete_inventory(inventory_id, user_id, retries=retries - 1)
 
 
 def with_path_cleanup(f):
@@ -650,8 +632,7 @@ def with_path_cleanup(f):
     return _wrapped
 
 
-class BaseTask(Task):
-    name = None
+class BaseTask(object):
     model = None
     event_model = None
     abstract = True
@@ -844,7 +825,12 @@ class BaseTask(Task):
         return False
 
     def build_inventory(self, instance, **kwargs):
-        json_data = json.dumps(instance.inventory.get_script_data(hostvars=True))
+        script_params = dict(hostvars=True)
+        if hasattr(instance, 'job_slice_number'):
+            script_params['slice_number'] = instance.job_slice_number
+            script_params['slice_count'] = instance.job_slice_count
+        script_data = instance.inventory.get_script_data(**script_params)
+        json_data = json.dumps(script_data)
         handle, path = tempfile.mkstemp(dir=kwargs.get('private_data_dir', None))
         f = os.fdopen(handle, 'w')
         f.write('#! /usr/bin/env python\n# -*- coding: utf-8 -*-\nprint %r\n' % json_data)
@@ -945,14 +931,11 @@ class BaseTask(Task):
             if instance.cancel_flag:
                 instance = self.update_model(instance.pk, status='canceled')
             if instance.status != 'running':
-                if hasattr(settings, 'CELERY_UNIT_TEST'):
-                    return
-                else:
-                    # Stop the task chain and prevent starting the job if it has
-                    # already been canceled.
-                    instance = self.update_model(pk)
-                    status = instance.status
-                    raise RuntimeError('not starting %s task' % instance.status)
+                # Stop the task chain and prevent starting the job if it has
+                # already been canceled.
+                instance = self.update_model(pk)
+                status = instance.status
+                raise RuntimeError('not starting %s task' % instance.status)
 
             if not os.path.exists(settings.AWX_PROOT_BASE_PATH):
                 raise RuntimeError('AWX_PROOT_BASE_PATH=%s does not exist' % settings.AWX_PROOT_BASE_PATH)
@@ -1085,8 +1068,6 @@ class BaseTask(Task):
             logger.exception(six.text_type('{} Final run hook errored.').format(instance.log_format))
         instance.websocket_emit_status(status)
         if status != 'successful':
-            # Raising an exception will mark the job as 'failed' in celery
-            # and will stop a task chain from continuing to execute
             if status == 'canceled':
                 raise AwxTaskError.TaskCancel(instance, rc)
             else:
@@ -1109,12 +1090,12 @@ class BaseTask(Task):
         return ''
 
 
+@task()
 class RunJob(BaseTask):
     '''
-    Celery task to run a job using ansible-playbook.
+    Run a job using ansible-playbook.
     '''
 
-    name = 'awx.main.tasks.run_job'
     model = Job
     event_model = JobEvent
     event_data_key = 'job_id'
@@ -1404,7 +1385,6 @@ class RunJob(BaseTask):
             self.update_model(job.pk, status='failed', job_explanation=error)
             raise RuntimeError(error)
         if job.project and job.project.scm_type:
-            job_request_id = '' if self.request.id is None else self.request.id
             pu_ig = job.instance_group
             pu_en = job.execution_node
             if job.is_isolated() is True:
@@ -1417,16 +1397,14 @@ class RunJob(BaseTask):
                     status='running',
                     instance_group = pu_ig,
                     execution_node=pu_en,
-                    celery_task_id=job_request_id))
+                    celery_task_id=job.celery_task_id))
             # save the associated job before calling run() so that a
             # cancel() call on the job can cancel the project update
             job = self.update_model(job.pk, project_update=local_project_sync)
 
             project_update_task = local_project_sync._get_task_class()
             try:
-                task_instance = project_update_task()
-                task_instance.request.id = job_request_id
-                task_instance.run(local_project_sync.id)
+                project_update_task().run(local_project_sync.id)
                 job = self.update_model(job.pk, scm_revision=job.project.scm_revision)
             except Exception:
                 local_project_sync.refresh_from_db()
@@ -1436,7 +1414,6 @@ class RunJob(BaseTask):
                                                              ('project_update', local_project_sync.name, local_project_sync.id)))
                     raise
 
-
     def final_run_hook(self, job, status, **kwargs):
         super(RunJob, self).final_run_hook(job, status, **kwargs)
         if job.use_fact_cache:
@@ -1467,9 +1444,9 @@ class RunJob(BaseTask):
             update_inventory_computed_fields.delay(inventory.id, True)
 
 
+@task()
 class RunProjectUpdate(BaseTask):
 
-    name = 'awx.main.tasks.run_project_update'
     model = ProjectUpdate
     event_model = ProjectUpdateEvent
     event_data_key = 'project_update_id'
@@ -1670,7 +1647,6 @@ class RunProjectUpdate(BaseTask):
         return getattr(settings, 'PROJECT_UPDATE_IDLE_TIMEOUT', None)
 
     def _update_dependent_inventories(self, project_update, dependent_inventory_sources):
-        project_request_id = '' if self.request.id is None else self.request.id
         scm_revision = project_update.project.scm_revision
         inv_update_class = InventoryUpdate._get_task_class()
         for inv_src in dependent_inventory_sources:
@@ -1693,13 +1669,10 @@ class RunProjectUpdate(BaseTask):
                         status='running',
                         instance_group=project_update.instance_group,
                         execution_node=project_update.execution_node,
-                        celery_task_id=str(project_request_id),
-                        source_project_update=project_update))
+                        source_project_update=project_update,
+                        celery_task_id=project_update.celery_task_id))
             try:
-                task_instance = inv_update_class()
-                # Runs in the same Celery task as project update
-                task_instance.request.id = project_request_id
-                task_instance.run(local_inv_update.id)
+                inv_update_class().run(local_inv_update.id)
             except Exception:
                 logger.exception(six.text_type('{} Unhandled exception updating dependent SCM inventory sources.')
                                  .format(project_update.log_format))
@@ -1804,9 +1777,9 @@ class RunProjectUpdate(BaseTask):
         return getattr(settings, 'AWX_PROOT_ENABLED', False)
 
 
+@task()
 class RunInventoryUpdate(BaseTask):
 
-    name = 'awx.main.tasks.run_inventory_update'
     model = InventoryUpdate
     event_model = InventoryUpdateEvent
     event_data_key = 'inventory_update_id'
@@ -2024,8 +1997,7 @@ class RunInventoryUpdate(BaseTask):
         This dictionary is used by `build_env`, below.
         """
         # Run the superclass implementation.
-        super_ = super(RunInventoryUpdate, self).build_passwords
-        passwords = super_(inventory_update, **kwargs)
+        passwords = super(RunInventoryUpdate, self).build_passwords(inventory_update, **kwargs)
 
         # Take key fields from the credential in use and add them to the
         # passwords dictionary.
@@ -2188,7 +2160,6 @@ class RunInventoryUpdate(BaseTask):
         if inventory_update.inventory_source:
             source_project = inventory_update.inventory_source.source_project
         if (inventory_update.source=='scm' and inventory_update.launch_type!='scm' and source_project):
-            request_id = '' if self.request.id is None else self.request.id
             local_project_sync = source_project.create_project_update(
                 _eager_fields=dict(
                     launch_type="sync",
@@ -2196,16 +2167,14 @@ class RunInventoryUpdate(BaseTask):
                     status='running',
                     execution_node=inventory_update.execution_node,
                     instance_group = inventory_update.instance_group,
-                    celery_task_id=request_id))
+                    celery_task_id=inventory_update.celery_task_id))
             # associate the inventory update before calling run() so that a
             # cancel() call on the inventory update can cancel the project update
             local_project_sync.scm_inventory_updates.add(inventory_update)
 
             project_update_task = local_project_sync._get_task_class()
             try:
-                task_instance = project_update_task()
-                task_instance.request.id = request_id
-                task_instance.run(local_project_sync.id)
+                project_update_task().run(local_project_sync.id)
                 inventory_update.inventory_source.scm_last_revision = local_project_sync.project.scm_revision
                 inventory_update.inventory_source.save(update_fields=['scm_last_revision'])
             except Exception:
@@ -2216,12 +2185,12 @@ class RunInventoryUpdate(BaseTask):
                 raise
 
 
+@task()
 class RunAdHocCommand(BaseTask):
     '''
-    Celery task to run an ad hoc command using ansible.
+    Run an ad hoc command using ansible.
     '''
 
-    name = 'awx.main.tasks.run_ad_hoc_command'
     model = AdHocCommand
     event_model = AdHocCommandEvent
     event_data_key = 'ad_hoc_command_id'
@@ -2382,9 +2351,9 @@ class RunAdHocCommand(BaseTask):
         return getattr(settings, 'AWX_PROOT_ENABLED', False)
 
 
+@task()
 class RunSystemJob(BaseTask):
 
-    name = 'awx.main.tasks.run_system_job'
     model = SystemJob
     event_model = SystemJobEvent
     event_data_key = 'system_job_id'
@@ -2439,9 +2408,9 @@ def _reconstruct_relationships(copy_mapping):
         new_obj.save()
 
 
-@shared_task(bind=True, queue=settings.CELERY_DEFAULT_QUEUE)
+@task()
 def deep_copy_model_obj(
-    self, model_module, model_name, obj_pk, new_obj_pk,
+    model_module, model_name, obj_pk, new_obj_pk,
     user_pk, sub_obj_list, permission_check_func=None
 ):
     logger.info(six.text_type('Deep copy {} from {} to {}.').format(model_name, obj_pk, new_obj_pk))

awx/main/tests/docs/test_swagger_generation.py

@@ -119,7 +119,7 @@ class TestSwaggerGeneration():
     def test_autogen_response_examples(self, swagger_autogen):
         for pattern, node in TestSwaggerGeneration.JSON['paths'].items():
             pattern = pattern.replace('{id}', '[0-9]+')
-            pattern = pattern.replace('{category_slug}', '[a-zA-Z0-9\-]+')
+            pattern = pattern.replace(r'{category_slug}', r'[a-zA-Z0-9\-]+')
             for path, result in swagger_autogen.items():
                 if re.match('^{}$'.format(pattern), path):
                     for key, value in result.items():

awx/main/tests/functional/api/test_credential_type.py

@@ -359,18 +359,17 @@ def test_create_with_valid_injectors(get, post, admin):
         },
         'injectors': {
             'env': {
-                'ANSIBLE_MY_CLOUD_TOKEN': '{{api_token}}'
+                'AWX_MY_CLOUD_TOKEN': '{{api_token}}'
             }
         }
-    }, admin)
-    assert response.status_code == 201
+    }, admin, expect=201)
 
     response = get(reverse('api:credential_type_list'), admin)
     assert response.data['count'] == 1
     injectors = response.data['results'][0]['injectors']
     assert len(injectors) == 1
     assert injectors['env'] == {
-        'ANSIBLE_MY_CLOUD_TOKEN': '{{api_token}}'
+        'AWX_MY_CLOUD_TOKEN': '{{api_token}}'
     }
 
 
@@ -388,7 +387,7 @@ def test_create_with_undefined_template_variable_xfail(post, admin):
             }]
         },
         'injectors': {
-            'env': {'ANSIBLE_MY_CLOUD_TOKEN': '{{api_tolkien}}'}
+            'env': {'AWX_MY_CLOUD_TOKEN': '{{api_tolkien}}'}
         }
     }, admin)
     assert response.status_code == 400

awx/main/tests/functional/api/test_job.py

@@ -122,6 +122,22 @@ def test_job_relaunch_on_failed_hosts(post, inventory, project, machine_credenti
     assert r.data.get('limit') == hosts
 
 
+@pytest.mark.django_db
+def test_slice_jt_recent_jobs(slice_job_factory, admin_user, get):
+    workflow_job = slice_job_factory(3, spawn=True)
+    slice_jt = workflow_job.job_template
+    r = get(
+        url=slice_jt.get_absolute_url(),
+        user=admin_user,
+        expect=200
+    )
+    job_ids = [entry['id'] for entry in r.data['summary_fields']['recent_jobs']]
+    assert workflow_job.pk not in job_ids
+    for node in workflow_job.workflow_nodes.all():
+        job = node.job
+        assert job.pk in job_ids
+
+
 @pytest.mark.django_db
 def test_block_unprocessed_events(delete, admin_user, mocker):
     time_of_finish = parse("Thu Feb 28 09:10:20 2013 -0500")

awx/main/tests/functional/api/test_job_runtime_params.py

@@ -6,7 +6,7 @@ import json
 from awx.api.serializers import JobLaunchSerializer
 from awx.main.models.credential import Credential
 from awx.main.models.inventory import Inventory, Host
-from awx.main.models.jobs import Job, JobTemplate
+from awx.main.models.jobs import Job, JobTemplate, UnifiedJobTemplate
 
 from awx.api.versioning import reverse
 
@@ -553,15 +553,15 @@ def test_callback_accept_prompted_extra_var(mocker, survey_spec_factory, job_tem
 
     with mocker.patch('awx.main.access.BaseAccess.check_license'):
         mock_job = mocker.MagicMock(spec=Job, id=968, extra_vars={"job_launch_var": 3, "survey_var": 4})
-        with mocker.patch.object(JobTemplate, 'create_unified_job', return_value=mock_job):
+        with mocker.patch.object(UnifiedJobTemplate, 'create_unified_job', return_value=mock_job):
             with mocker.patch('awx.api.serializers.JobSerializer.to_representation', return_value={}):
                 with mocker.patch('awx.api.views.JobTemplateCallback.find_matching_hosts', return_value=[host]):
                     post(
                         reverse('api:job_template_callback', kwargs={'pk': job_template.pk}),
                         dict(extra_vars={"job_launch_var": 3, "survey_var": 4}, host_config_key="foo"),
                         admin_user, expect=201, format='json')
-                    assert JobTemplate.create_unified_job.called
-                    assert JobTemplate.create_unified_job.call_args == ({
+                    assert UnifiedJobTemplate.create_unified_job.called
+                    assert UnifiedJobTemplate.create_unified_job.call_args == ({
                         'extra_vars': {'survey_var': 4, 'job_launch_var': 3},
                         '_eager_fields': {'launch_type': 'callback'},
                         'limit': 'single-host'},
@@ -579,15 +579,15 @@ def test_callback_ignore_unprompted_extra_var(mocker, survey_spec_factory, job_t
 
     with mocker.patch('awx.main.access.BaseAccess.check_license'):
         mock_job = mocker.MagicMock(spec=Job, id=968, extra_vars={"job_launch_var": 3, "survey_var": 4})
-        with mocker.patch.object(JobTemplate, 'create_unified_job', return_value=mock_job):
+        with mocker.patch.object(UnifiedJobTemplate, 'create_unified_job', return_value=mock_job):
             with mocker.patch('awx.api.serializers.JobSerializer.to_representation', return_value={}):
                 with mocker.patch('awx.api.views.JobTemplateCallback.find_matching_hosts', return_value=[host]):
                     post(
                         reverse('api:job_template_callback', kwargs={'pk':job_template.pk}),
                         dict(extra_vars={"job_launch_var": 3, "survey_var": 4}, host_config_key="foo"),
                         admin_user, expect=201, format='json')
-                    assert JobTemplate.create_unified_job.called
-                    assert JobTemplate.create_unified_job.call_args == ({
+                    assert UnifiedJobTemplate.create_unified_job.called
+                    assert UnifiedJobTemplate.create_unified_job.call_args == ({
                         '_eager_fields': {'launch_type': 'callback'},
                         'limit': 'single-host'},
                     )

awx/main/tests/functional/api/test_oauth.py

@@ -5,10 +5,7 @@ import json
 from django.db import connection
 from django.test.utils import override_settings
 from django.test import Client
-from django.core.urlresolvers import resolve
-from rest_framework.test import APIRequestFactory
 
-from awx.main.middleware import DeprecatedAuthTokenMiddleware
 from awx.main.utils.encryption import decrypt_value, get_encryption_key
 from awx.api.versioning import reverse, drf_reverse
 from awx.main.models.oauth import (OAuth2Application as Application, 
@@ -361,52 +358,3 @@ def test_revoke_refreshtoken(oauth_application, post, get, delete, admin):
     new_refresh_token = RefreshToken.objects.all().first()
     assert refresh_token == new_refresh_token
     assert new_refresh_token.revoked
-
-
-@pytest.mark.django_db
-@pytest.mark.parametrize('fmt', ['json', 'multipart'])
-def test_deprecated_authtoken_support(alice, fmt):
-    kwargs = {
-        'data': {'username': 'alice', 'password': 'alice'},
-        'format': fmt
-    }
-    request = getattr(APIRequestFactory(), 'post')('/api/v2/authtoken/', **kwargs)
-    DeprecatedAuthTokenMiddleware().process_request(request)
-    assert request.path == request.path_info == '/api/v2/users/{}/personal_tokens/'.format(alice.pk)
-    view, view_args, view_kwargs = resolve(request.path)
-    resp = view(request, *view_args, **view_kwargs)
-    assert resp.status_code == 201
-    assert 'token' in resp.data
-    assert resp.data['refresh_token'] is None
-    assert resp.data['scope'] == 'write'
-
-    for _type in ('Token', 'Bearer'):
-        request = getattr(APIRequestFactory(), 'get')(
-            '/api/v2/me/',
-            HTTP_AUTHORIZATION=' '.join([_type, resp.data['token']])
-        )
-        DeprecatedAuthTokenMiddleware().process_request(request)
-        view, view_args, view_kwargs = resolve(request.path)
-        assert view(request, *view_args, **view_kwargs).status_code == 200
-
-
-@pytest.mark.django_db
-def test_deprecated_authtoken_invalid_username(alice):
-    kwargs = {
-        'data': {'username': 'nobody', 'password': 'nobody'},
-        'format': 'json'
-    }
-    request = getattr(APIRequestFactory(), 'post')('/api/v2/authtoken/', **kwargs)
-    resp = DeprecatedAuthTokenMiddleware().process_request(request)
-    assert resp.status_code == 401
-
-
-@pytest.mark.django_db
-def test_deprecated_authtoken_missing_credentials(alice):
-    kwargs = {
-        'data': {},
-        'format': 'json'
-    }
-    request = getattr(APIRequestFactory(), 'post')('/api/v2/authtoken/', **kwargs)
-    resp = DeprecatedAuthTokenMiddleware().process_request(request)
-    assert resp.status_code == 401

awx/main/tests/functional/api/test_settings.py

@@ -7,6 +7,7 @@ import pytest
 import os
 
 from django.conf import settings
+from kombu.utils.url import parse_url
 
 # Mock
 import mock
@@ -358,3 +359,25 @@ def test_isolated_key_flag_readonly(get, patch, delete, admin):
 
     delete(url, user=admin)
     assert settings.AWX_ISOLATED_KEY_GENERATION is True
+
+
+@pytest.mark.django_db
+def test_default_broker_url():
+    url = parse_url(settings.BROKER_URL)
+    assert url['transport'] == 'amqp'
+    assert url['hostname'] == 'rabbitmq'
+    assert url['userid'] == 'guest'
+    assert url['password'] == 'guest'
+    assert url['virtual_host'] == '/'
+
+
+@pytest.mark.django_db
+def test_broker_url_with_special_characters():
+    settings.BROKER_URL = 'amqp://guest:a@ns:ibl3#@rabbitmq:5672//'
+    url = parse_url(settings.BROKER_URL)
+    assert url['transport'] == 'amqp'
+    assert url['hostname'] == 'rabbitmq'
+    assert url['port'] == 5672
+    assert url['userid'] == 'guest'
+    assert url['password'] == 'a@ns:ibl3#'
+    assert url['virtual_host'] == '/'

awx/main/tests/functional/api/test_survey_spec.py

@@ -285,9 +285,7 @@ def test_survey_spec_passwords_with_default_required(job_template_factory, post,
 @pytest.mark.django_db
 @pytest.mark.parametrize('default, status', [
     ('SUPERSECRET', 200),
-    (['some', 'invalid', 'list'], 400),
     ({'some-invalid': 'dict'}, 400),
-    (False, 400)
 ])
 def test_survey_spec_default_passwords_are_encrypted(job_template, post, admin_user, default, status):
     job_template.survey_enabled = True
@@ -317,7 +315,7 @@ def test_survey_spec_default_passwords_are_encrypted(job_template, post, admin_u
             'secret_value': default
         }
     else:
-        assert "for 'secret_value' expected to be a string." in str(resp.data)
+        assert "expected to be string." in str(resp.data)
 
 
 @mock.patch('awx.api.views.feature_enabled', lambda feature: True)
@@ -346,37 +344,6 @@ def test_survey_spec_default_passwords_encrypted_on_update(job_template, post, p
     assert updated_jt.survey_spec == JobTemplate.objects.get(pk=job_template.pk).survey_spec
 
 
-# Tests related to survey content validation
-@mock.patch('awx.api.views.feature_enabled', lambda feature: True)
-@pytest.mark.django_db
-@pytest.mark.survey
-def test_survey_spec_non_dict_error(deploy_jobtemplate, post, admin_user):
-    """When a question doesn't follow the standard format, verify error thrown."""
-    response = post(
-        url=reverse('api:job_template_survey_spec', kwargs={'pk': deploy_jobtemplate.id}),
-        data={
-            "description": "Email of the submitter",
-            "spec": ["What is your email?"], "name": "Email survey"
-        },
-        user=admin_user,
-        expect=400
-    )
-    assert response.data['error'] == "Survey question 0 is not a json object."
-
-
-@mock.patch('awx.api.views.feature_enabled', lambda feature: True)
-@pytest.mark.django_db
-@pytest.mark.survey
-def test_survey_spec_dual_names_error(survey_spec_factory, deploy_jobtemplate, post, user):
-    response = post(
-        url=reverse('api:job_template_survey_spec', kwargs={'pk': deploy_jobtemplate.id}),
-        data=survey_spec_factory(['submitter_email', 'submitter_email']),
-        user=user('admin', True),
-        expect=400
-    )
-    assert response.data['error'] == "'variable' 'submitter_email' duplicated in survey question 1."
-
-
 # Test actions that should be allowed with non-survey license
 @mock.patch('awx.main.access.BaseAccess.check_license', new=mock_no_surveys)
 @pytest.mark.django_db

awx/main/tests/functional/conftest.py

@@ -8,13 +8,13 @@ import tempfile
 import shutil
 from datetime import timedelta
 from six.moves import xrange
+from mock import PropertyMock
 
 # Django
 from django.core.urlresolvers import resolve
 from django.utils.six.moves.urllib.parse import urlparse
 from django.utils import timezone
 from django.contrib.auth.models import User
-from django.conf import settings
 from django.core.serializers.json import DjangoJSONEncoder
 from django.db.backends.sqlite3.base import SQLiteCursorWrapper
 from jsonbfield.fields import JSONField
@@ -66,17 +66,6 @@ def swagger_autogen(requests=__SWAGGER_REQUESTS__):
     return requests
 
 
-@pytest.fixture(scope="session", autouse=True)
-def celery_memory_broker():
-    '''
-    FIXME: Not sure how "far" just setting the BROKER_URL will get us.
-    We may need to incluence CELERY's configuration like we do in the old unit tests (see base.py)
-
-    Allows django signal code to execute without the need for redis
-    '''
-    settings.BROKER_URL='memory://localhost/'
-
-
 @pytest.fixture
 def user():
     def u(name, is_superuser=False):
@@ -682,6 +671,24 @@ def job_template_labels(organization, job_template):
     return job_template
 
 
+@pytest.fixture
+def jt_linked(job_template_factory, credential, net_credential, vault_credential):
+    '''
+    A job template with a reasonably complete set of related objects to
+    test RBAC and other functionality affected by related objects
+    '''
+    objects = job_template_factory(
+        'testJT', organization='org1', project='proj1', inventory='inventory1',
+        credential='cred1')
+    jt = objects.job_template
+    jt.credentials.add(vault_credential)
+    jt.save()
+    # Add AWS cloud credential and network credential
+    jt.credentials.add(credential)
+    jt.credentials.add(net_credential)
+    return jt
+
+
 @pytest.fixture
 def workflow_job_template(organization):
     wjt = WorkflowJobTemplate(name='test-workflow_job_template', organization=organization)
@@ -764,3 +771,42 @@ def sqlite_copy_expert(request):
     request.addfinalizer(lambda: delattr(SQLiteCursorWrapper, 'copy_expert'))
     return path
 
+
+@pytest.fixture
+def disable_database_settings(mocker):
+    m = mocker.patch('awx.conf.settings.SettingsWrapper.all_supported_settings', new_callable=PropertyMock)
+    m.return_value = []
+
+
+@pytest.fixture
+def slice_jt_factory(inventory):
+    def r(N, jt_kwargs=None):
+        for i in range(N):
+            inventory.hosts.create(name='foo{}'.format(i))
+        if not jt_kwargs:
+            jt_kwargs = {}
+        return JobTemplate.objects.create(
+            name='slice-jt-from-factory',
+            job_slice_count=N,
+            inventory=inventory,
+            **jt_kwargs
+        )
+    return r
+
+
+@pytest.fixture
+def slice_job_factory(slice_jt_factory):
+    def r(N, jt_kwargs=None, prompts=None, spawn=False):
+        slice_jt = slice_jt_factory(N, jt_kwargs=jt_kwargs)
+        if not prompts:
+            prompts = {}
+        slice_job = slice_jt.create_unified_job(**prompts)
+        if spawn:
+            for node in slice_job.workflow_nodes.all():
+                # does what the task manager does for spawning workflow jobs
+                kv = node.get_job_kwargs()
+                job = node.unified_job_template.create_unified_job(**kv)
+                node.job = job
+                node.save()
+        return slice_job
+    return r

awx/main/tests/functional/models/test_inventory.py

@@ -38,6 +38,33 @@ class TestInventoryScript:
             'remote_tower_id': host.id
         }
 
+    def test_slice_subset(self, inventory):
+        for i in range(3):
+            inventory.hosts.create(name='host{}'.format(i))
+        for i in range(3):
+            assert inventory.get_script_data(slice_number=i + 1, slice_count=3) == {
+                'all': {'hosts': ['host{}'.format(i)]}
+            }
+
+    def test_slice_subset_with_groups(self, inventory):
+        hosts = []
+        for i in range(3):
+            host = inventory.hosts.create(name='host{}'.format(i))
+            hosts.append(host)
+        g1 = inventory.groups.create(name='contains_all_hosts')
+        for host in hosts:
+            g1.hosts.add(host)
+        g2 = inventory.groups.create(name='contains_two_hosts')
+        for host in hosts[:2]:
+            g2.hosts.add(host)
+        for i in range(3):
+            expected_data = {
+                'contains_all_hosts': {'hosts': ['host{}'.format(i)], 'children': [], 'vars': {}},
+            }
+            if i < 2:
+                expected_data['contains_two_hosts'] = {'hosts': ['host{}'.format(i)], 'children': [], 'vars': {}}
+            assert inventory.get_script_data(slice_number=i + 1, slice_count=3) == expected_data
+
 
 @pytest.mark.django_db
 class TestActiveCount:

awx/main/tests/functional/models/test_job.py

@@ -1,7 +1,7 @@
 import pytest
 import six
 
-from awx.main.models import JobTemplate, Job, JobHostSummary
+from awx.main.models import JobTemplate, Job, JobHostSummary, WorkflowJob
 from crum import impersonate
 
 
@@ -81,3 +81,23 @@ def test_job_host_summary_representation(host):
     jhs = JobHostSummary.objects.get(pk=jhs.id)
     host.delete()
     assert 'N/A changed=1 dark=2 failures=3 ok=4 processed=5 skipped=6' == six.text_type(jhs)
+
+
+@pytest.mark.django_db
+class TestSlicingModels:
+
+    def test_slice_workflow_spawn(self, slice_jt_factory):
+        slice_jt = slice_jt_factory(3)
+        job = slice_jt.create_unified_job()
+        assert isinstance(job, WorkflowJob)
+        assert job.job_template == slice_jt
+        assert job.unified_job_template == slice_jt
+        assert job.workflow_nodes.count() == 3
+
+    def test_slices_with_JT_and_prompts(self, slice_job_factory):
+        job = slice_job_factory(3, jt_kwargs={'ask_limit_on_launch': True}, prompts={'limit': 'foobar'}, spawn=True)
+        assert job.launch_config.prompts_dict() == {'limit': 'foobar'}
+        for node in job.workflow_nodes.all():
+            assert node.limit is None  # data not saved in node prompts
+            job = node.job
+            assert job.limit == 'foobar'

awx/main/tests/functional/models/test_unified_job.py

@@ -1,13 +1,16 @@
 import itertools
 import pytest
-import mock
+import six
 
 # Django
 from django.contrib.contenttypes.models import ContentType
 
 # AWX
-from awx.main.models import UnifiedJobTemplate, Job, JobTemplate, WorkflowJobTemplate, Project, WorkflowJob, Schedule
-from awx.main.models.ha import InstanceGroup
+from awx.main.models import (
+    UnifiedJobTemplate, Job, JobTemplate, WorkflowJobTemplate,
+    Project, WorkflowJob, Schedule,
+    Credential
+)
 
 
 @pytest.mark.django_db
@@ -55,9 +58,7 @@ class TestCreateUnifiedJob:
         job_with_links.save()
         job_with_links.credentials.add(machine_credential)
         job_with_links.credentials.add(net_credential)
-        with mocker.patch('awx.main.models.unified_jobs.UnifiedJobTemplate._get_unified_job_field_names',
-                          return_value=['inventory', 'credential', 'limit']):
-            second_job = job_with_links.copy_unified_job()
+        second_job = job_with_links.copy_unified_job()
 
         # Check that job data matches the original variables
         assert second_job.credential == job_with_links.credential
@@ -65,47 +66,20 @@ class TestCreateUnifiedJob:
         assert second_job.limit == 'my_server'
         assert net_credential in second_job.credentials.all()
 
-
-@pytest.mark.django_db
-class TestIsolatedRuns:
-
-    def test_low_capacity_isolated_instance_selected(self):
-        ig = InstanceGroup.objects.create(name='tower')
-        iso_ig = InstanceGroup.objects.create(name='thepentagon', controller=ig)
-        iso_ig.instances.create(hostname='iso1', capacity=50)
-        i2 = iso_ig.instances.create(hostname='iso2', capacity=200)
-        job = Job.objects.create(
-            instance_group=iso_ig,
-            celery_task_id='something',
-        )
-
-        mock_async = mock.MagicMock()
-        success_callback = mock.MagicMock()
-        error_callback = mock.MagicMock()
-
-        class MockTaskClass:
-            apply_async = mock_async
-
-        with mock.patch.object(job, '_get_task_class') as task_class:
-            task_class.return_value = MockTaskClass
-            job.start_celery_task([], error_callback, success_callback, 'thepentagon')
-        mock_async.assert_called_with([job.id], [], 
-                                      link_error=error_callback, 
-                                      link=success_callback, 
-                                      queue='thepentagon',
-                                      task_id='something')
-
-        i2.capacity = 20
-        i2.save()
-
-        with mock.patch.object(job, '_get_task_class') as task_class:
-            task_class.return_value = MockTaskClass
-            job.start_celery_task([], error_callback, success_callback, 'thepentagon')
-        mock_async.assert_called_with([job.id], [], 
-                                      link_error=error_callback, 
-                                      link=success_callback, 
-                                      queue='thepentagon',
-                                      task_id='something')
+    def test_job_relaunch_modifed_jt(self, jt_linked):
+        # Replace all credentials with a new one of same type
+        new_creds = []
+        for cred in jt_linked.credentials.all():
+            new_creds.append(Credential.objects.create(
+                name=six.text_type(cred.name) + six.text_type('_new'),
+                credential_type=cred.credential_type,
+                inputs=cred.inputs
+            ))
+        job = jt_linked.create_unified_job()
+        jt_linked.credentials.clear()
+        jt_linked.credentials.add(*new_creds)
+        relaunched_job = job.copy_unified_job()
+        assert set(relaunched_job.credentials.all()) == set(new_creds)
 
 
 @pytest.mark.django_db
@@ -178,3 +152,50 @@ def test_event_processing_not_finished():
 def test_event_model_undefined():
     wj = WorkflowJob.objects.create(name='foobar', status='finished')
     assert wj.event_processing_finished
+
+
+@pytest.mark.django_db
+class TestTaskImpact:
+    @pytest.fixture
+    def job_host_limit(self, job_template, inventory):
+        def r(hosts, forks):
+            for i in range(hosts):
+                inventory.hosts.create(name='foo' + str(i))
+            job = Job.objects.create(
+                name='fake-job',
+                launch_type='workflow',
+                job_template=job_template,
+                inventory=inventory,
+                forks=forks
+            )
+            return job
+        return r
+
+    def test_limit_task_impact(self, job_host_limit):
+        job = job_host_limit(5, 2)
+        assert job.task_impact == 2 + 1  # forks becomes constraint
+
+    def test_host_task_impact(self, job_host_limit):
+        job = job_host_limit(3, 5)
+        assert job.task_impact == 3 + 1  # hosts becomes constraint
+
+    def test_shard_task_impact(self, slice_job_factory):
+        # factory creates on host per slice
+        workflow_job = slice_job_factory(3, jt_kwargs={'forks': 50}, spawn=True)
+        # arrange the jobs by their number
+        jobs = [None for i in range(3)]
+        for node in workflow_job.workflow_nodes.all():
+            jobs[node.job.job_slice_number - 1] = node.job
+        # Even distribution - all jobs run on 1 host
+        assert [
+            len(jobs[0].inventory.get_script_data(slice_number=i + 1, slice_count=3)['all']['hosts'])
+            for i in range(3)
+        ] == [1, 1, 1]
+        assert [job.task_impact for job in jobs] == [2, 2, 2]  # plus one base task impact
+        # Uneven distribution - first job takes the extra host
+        jobs[0].inventory.hosts.create(name='remainder_foo')
+        assert [
+            len(jobs[0].inventory.get_script_data(slice_number=i + 1, slice_count=3)['all']['hosts'])
+            for i in range(3)
+        ] == [2, 1, 1]
+        assert [job.task_impact for job in jobs] == [3, 2, 2]

awx/main/tests/functional/models/test_workflow.py

@@ -7,6 +7,7 @@ from awx.main.models.workflow import WorkflowJob, WorkflowJobNode, WorkflowJobTe
 from awx.main.models.jobs import JobTemplate, Job
 from awx.main.models.projects import ProjectUpdate
 from awx.main.scheduler.dag_workflow import WorkflowDAG
+from awx.api.versioning import reverse
 
 # Django
 from django.test import TransactionTestCase
@@ -196,9 +197,15 @@ class TestWorkflowJobTemplate:
         assert test_view.is_valid_relation(node_assoc, nodes[1]) == {'Error': 'Multiple parent relationship not allowed.'}
         # test mutex validation
         test_view.relationship = 'failure_nodes'
-        node_assoc_1 = WorkflowJobTemplateNode.objects.create(workflow_job_template=wfjt)
-        assert (test_view.is_valid_relation(nodes[2], node_assoc_1) ==
-                {'Error': 'Cannot associate failure_nodes when always_nodes have been associated.'})
+        
+    def test_always_success_failure_creation(self, wfjt, admin, get):
+        wfjt_node = wfjt.workflow_job_template_nodes.all()[1]
+        node = WorkflowJobTemplateNode.objects.create(workflow_job_template=wfjt)
+        wfjt_node.always_nodes.add(node)
+        assert len(node.get_parent_nodes()) == 1
+        url = reverse('api:workflow_job_template_node_list') + str(wfjt_node.id) + '/'
+        resp = get(url, admin)
+        assert node.id in resp.data['always_nodes']
 
     def test_wfjt_unique_together_with_org(self, organization):
         wfjt1 = WorkflowJobTemplate(name='foo', organization=organization)

awx/main/tests/functional/task_management/test_scheduler.py

@@ -1,19 +1,10 @@
 import pytest
 import mock
 import json
-from datetime import timedelta, datetime
-
-from django.core.cache import cache
-from django.utils.timezone import now as tz_now
+from datetime import timedelta
 
 from awx.main.scheduler import TaskManager
 from awx.main.utils import encrypt_field
-from awx.main.models import (
-    Job,
-    Instance,
-    WorkflowJob,
-)
-from awx.main.models.notifications import JobNotificationMixin
 
 
 @pytest.mark.django_db
@@ -245,140 +236,3 @@ def test_shared_dependencies_launch(default_instance_group, job_template_factory
     iu = [x for x in ii.inventory_updates.all()]
     assert len(pu) == 1
     assert len(iu) == 1
-
-
-@pytest.mark.django_db
-def test_cleanup_interval(mock_cache):
-    with mock.patch.multiple('awx.main.scheduler.task_manager.cache', get=mock_cache.get, set=mock_cache.set):
-        assert mock_cache.get('last_celery_task_cleanup') is None
-
-        TaskManager().cleanup_inconsistent_celery_tasks()
-        last_cleanup = mock_cache.get('last_celery_task_cleanup')
-        assert isinstance(last_cleanup, datetime)
-
-        TaskManager().cleanup_inconsistent_celery_tasks()
-        assert cache.get('last_celery_task_cleanup') == last_cleanup
-
-
-class TestReaper():
-    @pytest.fixture
-    def all_jobs(self, mocker):
-        now = tz_now()
-
-        Instance.objects.create(hostname='host1', capacity=100)
-        Instance.objects.create(hostname='host2', capacity=100)
-        Instance.objects.create(hostname='host3_split', capacity=100)
-        Instance.objects.create(hostname='host4_offline', capacity=0)
-
-        j1 = Job.objects.create(status='pending', execution_node='host1')
-        j2 = Job.objects.create(status='waiting', celery_task_id='considered_j2')
-        j3 = Job.objects.create(status='waiting', celery_task_id='considered_j3')
-        j3.modified = now - timedelta(seconds=60)
-        j3.save(update_fields=['modified'])
-        j4 = Job.objects.create(status='running', celery_task_id='considered_j4', execution_node='host1')
-        j5 = Job.objects.create(status='waiting', celery_task_id='reapable_j5')
-        j5.modified = now - timedelta(seconds=60)
-        j5.save(update_fields=['modified'])
-        j6 = Job.objects.create(status='waiting', celery_task_id='considered_j6')
-        j6.modified = now - timedelta(seconds=60)
-        j6.save(update_fields=['modified'])
-        j7 = Job.objects.create(status='running', celery_task_id='considered_j7', execution_node='host2')
-        j8 = Job.objects.create(status='running', celery_task_id='reapable_j7', execution_node='host2')
-        j9 = Job.objects.create(status='waiting', celery_task_id='reapable_j8')
-        j9.modified = now - timedelta(seconds=60)
-        j9.save(update_fields=['modified'])
-        j10 = Job.objects.create(status='running', celery_task_id='host3_j10', execution_node='host3_split')
-
-        j11 = Job.objects.create(status='running', celery_task_id='host4_j11', execution_node='host4_offline')
-
-        j12 = WorkflowJob.objects.create(status='running', celery_task_id='workflow_job', execution_node='host1')
-
-        js = [j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12]
-        for j in js:
-            j.save = mocker.Mock(wraps=j.save)
-            j.websocket_emit_status = mocker.Mock()
-        return js
-
-    @pytest.fixture
-    def considered_jobs(self, all_jobs):
-        return all_jobs[2:7] + [all_jobs[10]]
-
-    @pytest.fixture
-    def running_tasks(self, all_jobs):
-        return {
-            'host1': [all_jobs[3]],
-            'host2': [all_jobs[7], all_jobs[8]],
-            'host3_split': [all_jobs[9]],
-            'host4_offline': [all_jobs[10]],
-        }
-
-    @pytest.fixture
-    def waiting_tasks(self, all_jobs):
-        return [all_jobs[2], all_jobs[4], all_jobs[5], all_jobs[8]]
-
-    @pytest.fixture
-    def reapable_jobs(self, all_jobs):
-        return [all_jobs[4], all_jobs[7], all_jobs[10]]
-
-    @pytest.fixture
-    def unconsidered_jobs(self, all_jobs):
-        return all_jobs[0:1] + all_jobs[5:7]
-
-    @pytest.fixture
-    def active_tasks(self):
-        return ([], {
-            'host1': ['considered_j2', 'considered_j3', 'considered_j4',],
-            'host2': ['considered_j6', 'considered_j7'],
-        })
-
-    @pytest.mark.django_db
-    @mock.patch.object(JobNotificationMixin, 'send_notification_templates')
-    @mock.patch.object(TaskManager, 'get_active_tasks', lambda self: ([], []))
-    def test_cleanup_inconsistent_task(self, notify, active_tasks, considered_jobs, reapable_jobs, running_tasks, waiting_tasks, mocker, settings):
-        settings.AWX_INCONSISTENT_TASK_INTERVAL = 0
-        tm = TaskManager()
-
-        tm.get_running_tasks = mocker.Mock(return_value=(running_tasks, waiting_tasks))
-        tm.get_active_tasks = mocker.Mock(return_value=active_tasks)
-        
-        tm.cleanup_inconsistent_celery_tasks()
-        
-        for j in considered_jobs:
-            if j not in reapable_jobs:
-                j.save.assert_not_called()
-
-        assert notify.call_count == 4
-        notify.assert_has_calls([mock.call('failed') for j in reapable_jobs], any_order=True)
-
-        for j in reapable_jobs:
-            j.websocket_emit_status.assert_called_once_with('failed')
-            assert j.status == 'failed'
-            assert j.job_explanation == (
-                'Task was marked as running in Tower but was not present in the job queue, so it has been marked as failed.'
-            )
-
-    @pytest.mark.django_db
-    def test_get_running_tasks(self, all_jobs):
-        tm = TaskManager()
-
-        # Ensure the query grabs the expected jobs
-        execution_nodes_jobs, waiting_jobs = tm.get_running_tasks()
-        assert 'host1' in execution_nodes_jobs
-        assert 'host2' in execution_nodes_jobs
-        assert 'host3_split' in execution_nodes_jobs
-
-        assert all_jobs[3] in execution_nodes_jobs['host1']
-
-        assert all_jobs[6] in execution_nodes_jobs['host2']
-        assert all_jobs[7] in execution_nodes_jobs['host2']
-        
-        assert all_jobs[9] in execution_nodes_jobs['host3_split']
-
-        assert all_jobs[10] in execution_nodes_jobs['host4_offline']
-
-        assert all_jobs[11] not in execution_nodes_jobs['host1']
-
-        assert all_jobs[2] in waiting_jobs
-        assert all_jobs[4] in waiting_jobs
-        assert all_jobs[5] in waiting_jobs
-        assert all_jobs[8] in waiting_jobs

awx/main/tests/functional/test_dispatch.py

@@ -0,0 +1,387 @@
+import datetime
+import multiprocessing
+import random
+import signal
+import time
+
+from django.utils.timezone import now as tz_now
+import pytest
+
+from awx.main.models import Job, WorkflowJob, Instance
+from awx.main.dispatch import reaper
+from awx.main.dispatch.pool import PoolWorker, WorkerPool, AutoscalePool
+from awx.main.dispatch.publish import task
+from awx.main.dispatch.worker import BaseWorker, TaskWorker
+
+
+@task()
+def add(a, b):
+    return a + b
+
+
+class BaseTask(object):
+
+    def add(self, a, b):
+        return add(a, b)
+
+
+@task()
+class Adder(BaseTask):
+    def run(self, a, b):
+        return super(Adder, self).add(a, b)
+
+
+@task(queue='hard-math')
+def multiply(a, b):
+    return a * b
+
+
+class SimpleWorker(BaseWorker):
+
+    def perform_work(self, body, *args):
+        pass
+
+
+class ResultWriter(BaseWorker):
+
+    def perform_work(self, body, result_queue):
+        result_queue.put(body + '!!!')
+
+
+class SlowResultWriter(BaseWorker):
+
+    def perform_work(self, body, result_queue):
+        time.sleep(3)
+        super(SlowResultWriter, self).perform_work(body, result_queue)
+
+
+@pytest.mark.usefixtures("disable_database_settings")
+class TestPoolWorker:
+
+    def setup_method(self, test_method):
+        self.worker = PoolWorker(1000, self.tick, tuple())
+
+    def tick(self):
+        self.worker.finished.put(self.worker.queue.get()['uuid'])
+        time.sleep(.5)
+
+    def test_qsize(self):
+        assert self.worker.qsize == 0
+        for i in range(3):
+            self.worker.put({'task': 'abc123'})
+        assert self.worker.qsize == 3
+
+    def test_put(self):
+        assert len(self.worker.managed_tasks) == 0
+        assert self.worker.messages_finished == 0
+        self.worker.put({'task': 'abc123'})
+
+        assert len(self.worker.managed_tasks) == 1
+        assert self.worker.messages_sent == 1
+
+    def test_managed_tasks(self):
+        self.worker.put({'task': 'abc123'})
+        self.worker.calculate_managed_tasks()
+        assert len(self.worker.managed_tasks) == 1
+
+        self.tick()
+        self.worker.calculate_managed_tasks()
+        assert len(self.worker.managed_tasks) == 0
+
+    def test_current_task(self):
+        self.worker.put({'task': 'abc123'})
+        assert self.worker.current_task['task'] == 'abc123'
+
+    def test_quit(self):
+        self.worker.quit()
+        assert self.worker.queue.get() == 'QUIT'
+
+    def test_idle_busy(self):
+        assert self.worker.idle is True
+        assert self.worker.busy is False
+        self.worker.put({'task': 'abc123'})
+        assert self.worker.busy is True
+        assert self.worker.idle is False
+
+
+@pytest.mark.django_db
+class TestWorkerPool:
+
+    def setup_method(self, test_method):
+        self.pool = WorkerPool(min_workers=3)
+
+    def teardown_method(self, test_method):
+        self.pool.stop(signal.SIGTERM)
+
+    def test_worker(self):
+        self.pool.init_workers(SimpleWorker().work_loop)
+        assert len(self.pool) == 3
+        for worker in self.pool.workers:
+            assert worker.messages_sent == 0
+            assert worker.alive is True
+
+    def test_single_task(self):
+        self.pool.init_workers(SimpleWorker().work_loop)
+        self.pool.write(0, 'xyz')
+        assert self.pool.workers[0].messages_sent == 1  # worker at index 0 handled one task
+        assert self.pool.workers[1].messages_sent == 0
+        assert self.pool.workers[2].messages_sent == 0
+
+    def test_queue_preference(self):
+        self.pool.init_workers(SimpleWorker().work_loop)
+        self.pool.write(2, 'xyz')
+        assert self.pool.workers[0].messages_sent == 0
+        assert self.pool.workers[1].messages_sent == 0
+        assert self.pool.workers[2].messages_sent == 1  # worker at index 2 handled one task
+
+    def test_worker_processing(self):
+        result_queue = multiprocessing.Queue()
+        self.pool.init_workers(ResultWriter().work_loop, result_queue)
+        for i in range(10):
+            self.pool.write(
+                random.choice(range(len(self.pool))),
+                'Hello, Worker {}'.format(i)
+            )
+        all_messages = [result_queue.get(timeout=1) for i in range(10)]
+        all_messages.sort()
+        assert all_messages == [
+            'Hello, Worker {}!!!'.format(i)
+            for i in range(10)
+        ]
+
+        total_handled = sum([worker.messages_sent for worker in self.pool.workers])
+        assert total_handled == 10
+
+
+@pytest.mark.django_db
+class TestAutoScaling:
+
+    def setup_method(self, test_method):
+        self.pool = AutoscalePool(min_workers=2, max_workers=10)
+
+    def teardown_method(self, test_method):
+        self.pool.stop(signal.SIGTERM)
+
+    def test_scale_up(self):
+        result_queue = multiprocessing.Queue()
+        self.pool.init_workers(SlowResultWriter().work_loop, result_queue)
+
+        # start with two workers, write an event to each worker and make it busy
+        assert len(self.pool) == 2
+        for i, w in enumerate(self.pool.workers):
+            w.put('Hello, Worker {}'.format(0))
+        assert len(self.pool) == 2
+
+        # wait for the subprocesses to start working on their tasks and be marked busy
+        time.sleep(1)
+        assert self.pool.should_grow
+
+        # write a third message, expect a new worker to spawn because all
+        # workers are busy
+        self.pool.write(0, 'Hello, Worker {}'.format(2))
+        assert len(self.pool) == 3
+
+    def test_scale_down(self):
+        self.pool.init_workers(ResultWriter().work_loop, multiprocessing.Queue())
+
+        # start with two workers, and scale up to 10 workers
+        assert len(self.pool) == 2
+        for i in range(8):
+            self.pool.up()
+        assert len(self.pool) == 10
+
+        # cleanup should scale down to 8 workers
+        self.pool.cleanup()
+        assert len(self.pool) == 2
+
+    def test_max_scale_up(self):
+        self.pool.init_workers(ResultWriter().work_loop, multiprocessing.Queue())
+
+        assert len(self.pool) == 2
+        for i in range(25):
+            self.pool.up()
+        assert self.pool.max_workers == 10
+        assert self.pool.full is True
+        assert len(self.pool) == 10
+
+    def test_equal_worker_distribution(self):
+        # if all workers are busy, spawn new workers *before* adding messages
+        # to an existing queue
+        self.pool.init_workers(SlowResultWriter().work_loop, multiprocessing.Queue)
+
+        # start with two workers, write an event to each worker and make it busy
+        assert len(self.pool) == 2
+        for i in range(10):
+            self.pool.write(0, 'Hello, World!')
+        assert len(self.pool) == 10
+        for w in self.pool.workers:
+            assert w.busy
+            assert len(w.managed_tasks) == 1
+
+        # the queue is full at 10, the _next_ write should put the message into
+        # a worker's backlog
+        assert len(self.pool) == 10
+        for w in self.pool.workers:
+            assert w.messages_sent == 1
+        self.pool.write(0, 'Hello, World!')
+        assert len(self.pool) == 10
+        assert self.pool.workers[0].messages_sent == 2
+
+    def test_lost_worker_autoscale(self):
+        # if a worker exits, it should be replaced automatically up to min_workers
+        self.pool.init_workers(ResultWriter().work_loop, multiprocessing.Queue())
+
+        # start with two workers, kill one of them
+        assert len(self.pool) == 2
+        assert not self.pool.should_grow
+        alive_pid = self.pool.workers[1].pid
+        self.pool.workers[0].process.terminate()
+        time.sleep(1)  # wait a moment for sigterm
+
+        # clean up and the dead worker
+        self.pool.cleanup()
+        assert len(self.pool) == 1
+        assert self.pool.workers[0].pid == alive_pid
+
+        # the next queue write should replace the lost worker
+        self.pool.write(0, 'Hello, Worker')
+        assert len(self.pool) == 2
+
+
+@pytest.mark.usefixtures("disable_database_settings")
+class TestTaskDispatcher:
+
+    @property
+    def tm(self):
+        return TaskWorker()
+
+    def test_function_dispatch(self):
+        result = self.tm.perform_work({
+            'task': 'awx.main.tests.functional.test_dispatch.add',
+            'args': [2, 2]
+        })
+        assert result == 4
+
+    def test_method_dispatch(self):
+        result = self.tm.perform_work({
+            'task': 'awx.main.tests.functional.test_dispatch.Adder',
+            'args': [2, 2]
+        })
+        assert result == 4
+
+
+class TestTaskPublisher:
+
+    def test_function_callable(self):
+        assert add(2, 2) == 4
+
+    def test_method_callable(self):
+        assert Adder().run(2, 2) == 4
+
+    def test_function_apply_async(self):
+        message, queue = add.apply_async([2, 2])
+        assert message['args'] == [2, 2]
+        assert message['kwargs'] == {}
+        assert message['task'] == 'awx.main.tests.functional.test_dispatch.add'
+        assert queue == 'awx_private_queue'
+
+    def test_method_apply_async(self):
+        message, queue = Adder.apply_async([2, 2])
+        assert message['args'] == [2, 2]
+        assert message['kwargs'] == {}
+        assert message['task'] == 'awx.main.tests.functional.test_dispatch.Adder'
+        assert queue == 'awx_private_queue'
+
+    def test_apply_with_queue(self):
+        message, queue = add.apply_async([2, 2], queue='abc123')
+        assert queue == 'abc123'
+
+    def test_queue_defined_in_task_decorator(self):
+        message, queue = multiply.apply_async([2, 2])
+        assert queue == 'hard-math'
+
+    def test_queue_overridden_from_task_decorator(self):
+        message, queue = multiply.apply_async([2, 2], queue='not-so-hard')
+        assert queue == 'not-so-hard'
+
+    def test_apply_with_callable_queuename(self):
+        message, queue = add.apply_async([2, 2], queue=lambda: 'called')
+        assert queue == 'called'
+
+
+yesterday = tz_now() - datetime.timedelta(days=1)
+
+
+@pytest.mark.django_db
+class TestJobReaper(object):
+
+    @pytest.mark.parametrize('status, execution_node, controller_node, modified, fail', [
+        ('running', '', '', None, False),        # running, not assigned to the instance
+        ('running', 'awx', '', None, True),      # running, has the instance as its execution_node
+        ('running', '', 'awx', None, True),      # running, has the instance as its controller_node
+        ('waiting', '', '', None, False),        # waiting, not assigned to the instance
+        ('waiting', 'awx', '', None, False),     # waiting, was edited less than a minute ago
+        ('waiting', '', 'awx', None, False),     # waiting, was edited less than a minute ago
+        ('waiting', 'awx', '', yesterday, True), # waiting, assigned to the execution_node, stale
+        ('waiting', '', 'awx', yesterday, True), # waiting, assigned to the controller_node, stale
+    ])
+    def test_should_reap(self, status, fail, execution_node, controller_node, modified):
+        i = Instance(hostname='awx')
+        i.save()
+        j = Job(
+            status=status,
+            execution_node=execution_node,
+            controller_node=controller_node,
+            start_args='SENSITIVE',
+        )
+        j.save()
+        if modified:
+            # we have to edit the modification time _without_ calling save()
+            # (because .save() overwrites it to _now_)
+            Job.objects.filter(id=j.id).update(modified=modified)
+        reaper.reap(i)
+        job = Job.objects.first()
+        if fail:
+            assert job.status == 'failed'
+            assert 'marked as failed' in job.job_explanation
+            assert job.start_args == ''
+        else:
+            assert job.status == status
+
+    @pytest.mark.parametrize('excluded_uuids, fail', [
+        (['abc123'], False),
+        ([], True),
+    ])
+    def test_do_not_reap_excluded_uuids(self, excluded_uuids, fail):
+        i = Instance(hostname='awx')
+        i.save()
+        j = Job(
+            status='running',
+            execution_node='awx',
+            controller_node='',
+            start_args='SENSITIVE',
+            celery_task_id='abc123',
+        )
+        j.save()
+
+        # if the UUID is excluded, don't reap it
+        reaper.reap(i, excluded_uuids=excluded_uuids)
+        job = Job.objects.first()
+        if fail:
+            assert job.status == 'failed'
+            assert 'marked as failed' in job.job_explanation
+            assert job.start_args == ''
+        else:
+            assert job.status == 'running'
+
+    def test_workflow_does_not_reap(self):
+        i = Instance(hostname='awx')
+        i.save()
+        j = WorkflowJob(
+            status='running',
+            execution_node='awx'
+        )
+        j.save()
+        reaper.reap(i)
+
+        assert WorkflowJob.objects.first().status == 'running'

awx/main/tests/functional/test_rbac_job_start.py

@@ -3,6 +3,11 @@ import pytest
 from awx.main.models.inventory import Inventory
 from awx.main.models.credential import Credential
 from awx.main.models.jobs import JobTemplate, Job
+from awx.main.access import (
+    UnifiedJobAccess,
+    WorkflowJobAccess, WorkflowJobNodeAccess,
+    JobAccess
+)
 
 
 @pytest.mark.django_db
@@ -43,6 +48,31 @@ def test_inventory_use_access(inventory, user):
     assert common_user.can_access(Inventory, 'use', inventory)
 
 
+@pytest.mark.django_db
+def test_slice_job(slice_job_factory, rando):
+    workflow_job = slice_job_factory(2, jt_kwargs={'created_by': rando}, spawn=True)
+    workflow_job.job_template.execute_role.members.add(rando)
+
+    # Abilities of user with execute_role for slice workflow job container
+    assert WorkflowJobAccess(rando).can_start(workflow_job)  # relaunch allowed
+    for access_cls in (UnifiedJobAccess, WorkflowJobAccess):
+        access = access_cls(rando)
+        assert access.can_read(workflow_job)
+        assert workflow_job in access.get_queryset()
+
+    # Abilities of user with execute_role for all the slice of the job
+    for node in workflow_job.workflow_nodes.all():
+        access = WorkflowJobNodeAccess(rando)
+        assert access.can_read(node)
+        assert node in access.get_queryset()
+        job = node.job
+        assert JobAccess(rando).can_start(job)  # relaunch allowed
+        for access_cls in (UnifiedJobAccess, JobAccess):
+            access = access_cls(rando)
+            assert access.can_read(job)
+            assert job in access.get_queryset()
+
+
 @pytest.mark.django_db
 class TestJobRelaunchAccess:
     @pytest.fixture

awx/main/tests/functional/test_rbac_job_templates.py

@@ -15,24 +15,6 @@ from awx.main.models.organization import Organization
 from awx.main.models.schedules import Schedule
 
 
-@pytest.fixture
-def jt_linked(job_template_factory, credential, net_credential, vault_credential):
-    '''
-    A job template with a reasonably complete set of related objects to
-    test RBAC and other functionality affected by related objects
-    '''
-    objects = job_template_factory(
-        'testJT', organization='org1', project='proj1', inventory='inventory1',
-        credential='cred1')
-    jt = objects.job_template
-    jt.credentials.add(vault_credential)
-    jt.save()
-    # Add AWS cloud credential and network credential
-    jt.credentials.add(credential)
-    jt.credentials.add(net_credential)
-    return jt
-
-
 @mock.patch.object(BaseAccess, 'check_license', return_value=None)
 @pytest.mark.django_db
 def test_job_template_access_superuser(check_license, user, deploy_jobtemplate):

awx/main/tests/unit/api/test_views.py

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 import mock
 import pytest
 import requests
@@ -293,7 +294,7 @@ class TestSurveySpecValidation:
         new['spec'][0]['default'] = '$encrypted$'
         new['spec'][0]['required'] = False
         resp = view._validate_spec_data(new, old)
-        assert resp is None
+        assert resp is None, resp.data
         assert new == {
             "name": "old survey",
             "description": "foobar",
@@ -357,3 +358,58 @@ class TestSurveySpecValidation:
                 }
             ]
         }
+
+
+    @staticmethod
+    def spec_from_element(survey_item):
+        survey_item.setdefault('name', 'foo')
+        survey_item.setdefault('variable', 'foo')
+        survey_item.setdefault('required', False)
+        survey_item.setdefault('question_name', 'foo')
+        survey_item.setdefault('type', 'text')
+        spec = {
+            'name': 'test survey',
+            'description': 'foo',
+            'spec': [survey_item]
+        }
+        return spec
+
+
+    @pytest.mark.parametrize("survey_item, error_text", [
+        ({'type': 'password', 'default': ['some', 'invalid', 'list']}, 'expected to be string'),
+        ({'type': 'password', 'default': False}, 'expected to be string'),
+        ({'type': 'integer', 'default': 'foo'}, 'expected to be int'),
+        ({'type': 'integer', 'default': u'🐉'}, 'expected to be int'),
+        ({'type': 'foo'}, 'allowed question types'),
+        ({'type': u'🐉'}, 'allowed question types'),
+        ({'type': 'multiplechoice'}, 'multiplechoice must specify choices'),
+        ({'type': 'integer', 'min': 'foo'}, 'min limit in survey question 0 expected to be integer'),
+        ({'question_name': 42}, "'question_name' in survey question 0 expected to be string.")
+    ])
+    def test_survey_question_element_validation(self, survey_item, error_text):
+        spec = self.spec_from_element(survey_item)
+        r = JobTemplateSurveySpec._validate_spec_data(spec, {})
+        assert r is not None, (spec, error_text)
+        assert 'error' in r.data
+        assert error_text in r.data['error']
+
+
+    def test_survey_spec_non_dict_error(self):
+        spec = self.spec_from_element({})
+        spec['spec'][0] = 'foo'
+        r = JobTemplateSurveySpec._validate_spec_data(spec, {})
+        assert 'Survey question 0 is not a json object' in r.data['error']
+
+
+    def test_survey_spec_dual_names_error(self):
+        spec = self.spec_from_element({})
+        spec['spec'].append(spec['spec'][0].copy())
+        r = JobTemplateSurveySpec._validate_spec_data(spec, {})
+        assert "'variable' 'foo' duplicated in survey question 1." in r.data['error']
+
+
+    def test_survey_spec_element_missing_property(self):
+        spec = self.spec_from_element({})
+        spec['spec'][0].pop('type')
+        r = JobTemplateSurveySpec._validate_spec_data(spec, {})
+        assert "'type' missing from survey question 0" in r.data['error']

awx/main/tests/unit/models/test_inventory.py

@@ -73,17 +73,6 @@ def test_invalid_kind_clean_insights_credential():
 
 
 class TestControlledBySCM(): 
-    @pytest.mark.parametrize('source', [
-        'scm',
-        'ec2',
-        'manual',
-    ])
-    def test_clean_overwrite_vars_valid(self, source):
-        inv_src = InventorySource(overwrite_vars=True,
-                                  source=source)
-
-        inv_src.clean_overwrite_vars()
-
     def test_clean_source_path_valid(self):
         inv_src = InventorySource(source_path='/not_real/',
                                   source='scm')

awx/main/tests/unit/test_fields.py

@@ -1,4 +1,6 @@
+# -*- coding: utf-8 -*-
 import pytest
+import six
 
 from django.core.exceptions import ValidationError
 from rest_framework.serializers import ValidationError as DRFValidationError
@@ -123,6 +125,9 @@ def test_cred_type_input_schema_validity(input_, valid):
     ({'env': {'AWX_SECRET_99': '{{awx_secret}}'}}, True),
     ({'env': {'99': '{{awx_secret}}'}}, False),
     ({'env': {'AWX_SECRET=': '{{awx_secret}}'}}, False),
+    ({'env': {'ANSIBLE_SETTING': '{{awx_secret}}'}}, False),
+    ({'env': {'DRAGON': u'🐉'}}, False),
+    ({'env': {u'🐉': 'DRAGON'}}, False),
     ({'extra_vars': 123}, False),
     ({'extra_vars': {}}, True),
     ({'extra_vars': {'hostname': '{{host}}'}}, True),
@@ -147,7 +152,8 @@ def test_cred_type_injectors_schema(injectors, valid):
     )
     field = CredentialType._meta.get_field('injectors')
     if valid is False:
-        with pytest.raises(ValidationError):
+        with pytest.raises(ValidationError, message=six.text_type(
+                "Injector was supposed to throw a validation error, data: {}").format(injectors)):
             field.clean(injectors, type_)
     else:
         field.clean(injectors, type_)

awx/main/tests/unit/test_task_manager.py

@@ -1,69 +0,0 @@
-# Copyright (c) 2017 Ansible by Red Hat
-# All Rights Reserved.
-
-import mock
-import pytest
-
-from django.utils.timezone import now as tz_now
-from django.db import DatabaseError
-
-from awx.main.scheduler import TaskManager
-from awx.main.models import (
-    Job,
-    Instance,
-    InstanceGroup,
-)
-from django.core.cache import cache
-
-
-class TestCleanupInconsistentCeleryTasks():
-    @mock.patch.object(cache, 'get', return_value=None)
-    @mock.patch.object(TaskManager, 'get_active_tasks', return_value=([], {}))
-    @mock.patch.object(TaskManager, 'get_running_tasks', return_value=({'host1': [Job(id=2), Job(id=3),]}, []))
-    @mock.patch.object(InstanceGroup.objects, 'prefetch_related', return_value=[])
-    @mock.patch.object(Instance.objects, 'filter', return_value=mock.MagicMock(first=lambda: None))
-    @mock.patch('awx.main.scheduler.task_manager.logger')
-    def test_instance_does_not_exist(self, logger_mock, *args):
-        logger_mock.error = mock.MagicMock(side_effect=RuntimeError("mocked"))
-        tm = TaskManager()
-        with pytest.raises(RuntimeError) as excinfo:
-            tm.cleanup_inconsistent_celery_tasks()
-
-        assert "mocked" in str(excinfo.value)
-        logger_mock.error.assert_called_once_with("Execution node Instance host1 not found in database. "
-                                                  "The node is currently executing jobs ['job 2 (new)', "
-                                                  "'job 3 (new)']")
-
-    @mock.patch.object(cache, 'get', return_value=None)
-    @mock.patch.object(TaskManager, 'get_active_tasks', return_value=([], {'host1': []}))
-    @mock.patch.object(InstanceGroup.objects, 'prefetch_related', return_value=[])
-    @mock.patch.object(TaskManager, 'get_running_tasks')
-    @mock.patch('awx.main.scheduler.task_manager.logger')
-    def test_save_failed(self, logger_mock, get_running_tasks, *args):
-        logger_mock.error = mock.MagicMock()
-        job = Job(id=2, modified=tz_now(), status='running', celery_task_id='blah', execution_node='host1')
-        job.websocket_emit_status = mock.MagicMock()
-        get_running_tasks.return_value = ({'host1': [job]}, [])
-        tm = TaskManager()
-
-        with mock.patch.object(job, 'save', side_effect=DatabaseError):
-            tm.cleanup_inconsistent_celery_tasks()
-            job.save.assert_called_once()
-            logger_mock.error.assert_called_once_with("Task job 2 (failed) DB error in marking failed. Job possibly deleted.")
-
-    @mock.patch.object(InstanceGroup.objects, 'prefetch_related', return_value=[])
-    @mock.patch('awx.main.scheduler.task_manager.Inspect')
-    def test_multiple_active_instances_sanity_check(self, inspect_mock, *args):
-        class MockInspector:
-            pass
-
-        mock_inspector = MockInspector()
-        mock_inspector.active = lambda: {
-            'celery@host1': [],
-            'celery@host2': []
-        }
-        inspect_mock.return_value = mock_inspector
-        tm = TaskManager()
-        active_task_queues, queues = tm.get_active_tasks()
-        assert 'host1' in queues
-        assert 'host2' in queues

awx/main/tests/unit/test_tasks.py

@@ -67,7 +67,7 @@ def test_work_success_callback_missing_job():
     task_data = {'type': 'project_update', 'id': 9999}
     with mock.patch('django.db.models.query.QuerySet.get') as get_mock:
         get_mock.side_effect = ProjectUpdate.DoesNotExist()
-        assert tasks.handle_work_success(None, task_data) is None
+        assert tasks.handle_work_success(task_data) is None
 
 
 def test_send_notifications_list(mocker):
@@ -246,6 +246,8 @@ class TestJobExecution(object):
             # If `Job.update_model` is called, we're not actually persisting
             # to the database; just update the status, which is usually
             # the update we care about for testing purposes
+            if kwargs.get('result_traceback'):
+                raise Exception('Task encountered error:\n{}'.format(kwargs['result_traceback']))
             if 'status' in kwargs:
                 self.instance.status = kwargs['status']
             if 'job_env' in kwargs:
@@ -891,9 +893,9 @@ class TestJobCredentials(TestJobExecution):
             (k.pattern, v) for k, v in call_kwargs['expect_passwords'].items()
             if 'Vault' in k.pattern
         )
-        assert vault_passwords['Vault password \(prod\):\\s*?$'] == 'pass@prod'
-        assert vault_passwords['Vault password \(dev\):\\s*?$'] == 'pass@dev'
-        assert vault_passwords['Vault password:\\s*?$'] == ''
+        assert vault_passwords['Vault password \(prod\):\\s*?$'] == 'pass@prod'  # noqa
+        assert vault_passwords['Vault password \(dev\):\\s*?$'] == 'pass@dev'  # noqa
+        assert vault_passwords['Vault password:\\s*?$'] == ''  # noqa
         assert '--ask-vault-pass' not in ' '.join(args)
         assert '--vault-id dev@prompt' in ' '.join(args)
         assert '--vault-id prod@prompt' in ' '.join(args)
@@ -935,9 +937,9 @@ class TestJobCredentials(TestJobExecution):
             (k.pattern, v) for k, v in call_kwargs['expect_passwords'].items()
             if 'Vault' in k.pattern
         )
-        assert vault_passwords['Vault password \(prod\):\\s*?$'] == 'provided-at-launch@prod'
-        assert vault_passwords['Vault password \(dev\):\\s*?$'] == 'provided-at-launch@dev'
-        assert vault_passwords['Vault password:\\s*?$'] == ''
+        assert vault_passwords['Vault password \(prod\):\\s*?$'] == 'provided-at-launch@prod'  # noqa
+        assert vault_passwords['Vault password \(dev\):\\s*?$'] == 'provided-at-launch@dev'  # noqa
+        assert vault_passwords['Vault password:\\s*?$'] == ''  # noqa
         assert '--ask-vault-pass' not in ' '.join(args)
         assert '--vault-id dev@prompt' in ' '.join(args)
         assert '--vault-id prod@prompt' in ' '.join(args)

awx/main/tests/unit/test_views.py

@@ -8,6 +8,7 @@ from rest_framework.generics import ListAPIView
 # AWX
 from awx.main.views import ApiErrorView
 from awx.api.views import JobList, InventorySourceList
+from awx.api.generics import ListCreateAPIView, SubListAttachDetachAPIView
 
 
 HTTP_METHOD_NAMES = [
@@ -73,3 +74,30 @@ def test_views_have_search_fields(all_views):
                 for v in views_missing_search
             ]))
         )
+
+
+def test_global_creation_always_possible(all_views):
+    """To not make life very difficult for clients, this test
+    asserts that all creatable resources can be created by
+    POSTing to the global resource list
+    """
+    views_by_model = {}
+    for View in all_views:
+        if not getattr(View, 'deprecated', False) and issubclass(View, ListAPIView) and hasattr(View, 'model'):
+            views_by_model.setdefault(View.model, []).append(View)
+    for model, views in views_by_model.items():
+        creatable = False
+        global_view = None
+        creatable_view = None
+        for View in views:
+            if '{}ListView'.format(model.__name__) == View.__name__:
+                global_view = View
+            if issubclass(View, ListCreateAPIView) and not issubclass(View, SubListAttachDetachAPIView):
+                creatable = True
+                creatable_view = View
+        if not creatable or not global_view:
+            continue
+        assert 'POST' in global_view().allowed_methods, (
+            'Resource {} should be creatable in global list view {}. '
+            'Can be created now in {}'.format(model, global_view, creatable_view)
+        )

awx/main/tests/unit/utils/test_reload.py

@@ -8,8 +8,8 @@ def test_produce_supervisor_command(mocker):
     mock_process.communicate = communicate_mock
     Popen_mock = mocker.MagicMock(return_value=mock_process)
     with mocker.patch.object(reload.subprocess, 'Popen', Popen_mock):
-        reload._supervisor_service_command(['beat', 'callback', 'fact'], "restart")
+        reload._supervisor_service_command("restart")
         reload.subprocess.Popen.assert_called_once_with(
-            ['supervisorctl', 'restart', 'tower-processes:receiver',],
+            ['supervisorctl', 'restart', 'tower-processes:*',],
             stderr=-1, stdin=-1, stdout=-1)
 

awx/main/utils/autoscale.py

@@ -1,27 +0,0 @@
-from celery.utils.log import get_logger
-from celery.worker.autoscale import Autoscaler, AUTOSCALE_KEEPALIVE
-from django.conf import settings
-import psutil
-
-logger = get_logger('awx.main.tasks')
-
-
-class DynamicAutoScaler(Autoscaler):
-
-    def __init__(self, pool, max_concurrency, min_concurrency=0, worker=None,
-                 keepalive=AUTOSCALE_KEEPALIVE, mutex=None):
-        super(DynamicAutoScaler, self).__init__(pool, max_concurrency,
-                                                min_concurrency, worker,
-                                                keepalive, mutex)
-        settings_absmem = getattr(settings, 'SYSTEM_TASK_ABS_MEM', None)
-        if settings_absmem is not None:
-            total_memory_gb = int(settings_absmem)
-        else:
-            total_memory_gb = (psutil.virtual_memory().total >> 30) + 1  # noqa: round up
-
-        # 5 workers per GB of total memory
-        self.max_concurrency = min(max_concurrency, (total_memory_gb * 5))
-        logger.warn('celery worker dynamic --autoscale={},{}'.format(
-            self.max_concurrency,
-            self.min_concurrency
-        ))

awx/main/utils/filters.py

@@ -118,10 +118,10 @@ def string_to_type(t):
     elif t == u'false':
         return False
 
-    if re.search('^[-+]?[0-9]+$',t):
+    if re.search(r'^[-+]?[0-9]+$',t):
         return int(t)
 
-    if re.search('^[-+]?[0-9]+\.[0-9]+$',t):
+    if re.search(r'^[-+]?[0-9]+\.[0-9]+$',t):
         return float(t)
 
     return t

awx/main/utils/ha.py

@@ -1,17 +0,0 @@
-# -*- coding: utf-8 -*-
-
-# Copyright (c) 2017 Ansible Tower by Red Hat
-# All Rights Reserved.
-
-from django.conf import settings
-
-
-class AWXCeleryRouter(object):
-    def route_for_task(self, task, args=None, kwargs=None):
-        tasks = [
-            'awx.main.tasks.cluster_node_heartbeat',
-            'awx.main.tasks.purge_old_stdout_files',
-            'awx.main.tasks.awx_isolated_heartbeat',
-        ]
-        if task in tasks:
-            return {'queue': settings.CLUSTER_HOST_ID, 'routing_key': settings.CLUSTER_HOST_ID}

awx/main/utils/reload.py

@@ -11,11 +11,8 @@ from django.conf import settings
 logger = logging.getLogger('awx.main.utils.reload')
 
 
-def _supervisor_service_command(service_internal_names, command, communicate=True):
+def _supervisor_service_command(command, communicate=True):
     '''
-    Service internal name options:
-     - beat - celery - callback - channels - uwsgi - daphne
-     - fact - nginx
     example use pattern of supervisorctl:
     # supervisorctl restart tower-processes:receiver tower-processes:factcacher
     '''
@@ -25,13 +22,7 @@ def _supervisor_service_command(service_internal_names, command, communicate=Tru
     args = ['supervisorctl']
     if settings.DEBUG:
         args.extend(['-c', '/supervisor.conf'])
-    programs = []
-    name_translation_dict = settings.SERVICE_NAME_DICT
-    for n in service_internal_names:
-        if n in name_translation_dict:
-            programs.append('{}:{}'.format(group_name, name_translation_dict[n]))
-    args.extend([command])
-    args.extend(programs)
+    args.extend([command, '{}:*'.format(group_name)])
     logger.debug('Issuing command to {} services, args={}'.format(command, args))
     supervisor_process = subprocess.Popen(args, stdin=subprocess.PIPE,
                                           stdout=subprocess.PIPE, stderr=subprocess.PIPE)
@@ -48,6 +39,6 @@ def _supervisor_service_command(service_internal_names, command, communicate=Tru
         logger.info('Submitted supervisorctl {} command, not waiting for result'.format(command))
 
 
-def stop_local_services(service_internal_names, communicate=True):
-    logger.warn('Stopping services {} on this node in response to user action'.format(service_internal_names))
-    _supervisor_service_command(service_internal_names, command='stop', communicate=communicate)
+def stop_local_services(communicate=True):
+    logger.warn('Stopping services on this node in response to user action')
+    _supervisor_service_command(command='stop', communicate=communicate)

awx/main/utils/safe_yaml.py

@@ -78,10 +78,10 @@ def sanitize_jinja(arg):
     if isinstance(arg, six.string_types):
         # If the argument looks like it contains Jinja expressions
         # {{ x }} ...
-        if re.search('\{\{[^}]+}}', arg) is not None:
+        if re.search(r'\{\{[^}]+}}', arg) is not None:
             raise ValueError('Inline Jinja variables are not allowed.')
         # If the argument looks like it contains Jinja statements/control flow...
         # {% if x.foo() %} ...
-        if re.search('\{%[^%]+%}', arg) is not None:
+        if re.search(r'\{%[^%]+%}', arg) is not None:
             raise ValueError('Inline Jinja variables are not allowed.')
     return arg

awx/plugins/inventory/azure_rm.py

@@ -24,7 +24,7 @@ Azure External Inventory Script
 ===============================
 Generates dynamic inventory by making API requests to the Azure Resource
 Manager using the Azure Python SDK. For instruction on installing the
-Azure Python SDK see http://azure-sdk-for-python.readthedocs.org/
+Azure Python SDK see https://azure-sdk-for-python.readthedocs.io/
 
 Authentication
 --------------
@@ -50,6 +50,7 @@ Command line arguments:
  - ad_user
  - password
  - cloud_environment
+ - adfs_authority_url
 
 Environment variables:
  - AZURE_PROFILE
@@ -60,6 +61,7 @@ Environment variables:
  - AZURE_AD_USER
  - AZURE_PASSWORD
  - AZURE_CLOUD_ENVIRONMENT
+ - AZURE_ADFS_AUTHORITY_URL
 
 Run for Specific Host
 -----------------------
@@ -200,27 +202,43 @@ except ImportError:
     # python3
     import configparser as cp
 
-from packaging.version import Version
-
 from os.path import expanduser
 import ansible.module_utils.six.moves.urllib.parse as urlparse
 
 HAS_AZURE = True
 HAS_AZURE_EXC = None
+HAS_AZURE_CLI_CORE = True
+CLIError = None
 
 try:
+    from msrestazure.azure_active_directory import AADTokenCredentials
     from msrestazure.azure_exceptions import CloudError
+    from msrestazure.azure_active_directory import MSIAuthentication
     from msrestazure import azure_cloud
     from azure.mgmt.compute import __version__ as azure_compute_version
     from azure.common import AzureMissingResourceHttpError, AzureHttpError
     from azure.common.credentials import ServicePrincipalCredentials, UserPassCredentials
     from azure.mgmt.network import NetworkManagementClient
     from azure.mgmt.resource.resources import ResourceManagementClient
+    from azure.mgmt.resource.subscriptions import SubscriptionClient
     from azure.mgmt.compute import ComputeManagementClient
+    from adal.authentication_context import AuthenticationContext
 except ImportError as exc:
     HAS_AZURE_EXC = exc
     HAS_AZURE = False
 
+try:
+    from azure.cli.core.util import CLIError
+    from azure.common.credentials import get_azure_cli_credentials, get_cli_profile
+    from azure.common.cloud import get_cli_active_cloud
+except ImportError:
+    HAS_AZURE_CLI_CORE = False
+    CLIError = Exception
+
+try:
+    from ansible.release import __version__ as ansible_version
+except ImportError:
+    ansible_version = 'unknown'
 
 AZURE_CREDENTIAL_ENV_MAPPING = dict(
     profile='AZURE_PROFILE',
@@ -231,6 +249,7 @@ AZURE_CREDENTIAL_ENV_MAPPING = dict(
     ad_user='AZURE_AD_USER',
     password='AZURE_PASSWORD',
     cloud_environment='AZURE_CLOUD_ENVIRONMENT',
+    adfs_authority_url='AZURE_ADFS_AUTHORITY_URL'
 )
 
 AZURE_CONFIG_SETTINGS = dict(
@@ -241,10 +260,13 @@ AZURE_CONFIG_SETTINGS = dict(
     group_by_resource_group='AZURE_GROUP_BY_RESOURCE_GROUP',
     group_by_location='AZURE_GROUP_BY_LOCATION',
     group_by_security_group='AZURE_GROUP_BY_SECURITY_GROUP',
-    group_by_tag='AZURE_GROUP_BY_TAG'
+    group_by_tag='AZURE_GROUP_BY_TAG',
+    group_by_os_family='AZURE_GROUP_BY_OS_FAMILY',
+    use_private_ip='AZURE_USE_PRIVATE_IP'
 )
 
 AZURE_MIN_VERSION = "2.0.0"
+ANSIBLE_USER_AGENT = 'Ansible/{0}'.format(ansible_version)
 
 
 def azure_id_to_dict(id):
@@ -265,6 +287,8 @@ class AzureRM(object):
         self._compute_client = None
         self._resource_client = None
         self._network_client = None
+        self._adfs_authority_url = None
+        self._resource = None
 
         self.debug = False
         if args.debug:
@@ -300,13 +324,38 @@ class AzureRM(object):
         self.log("setting subscription_id")
         self.subscription_id = self.credentials['subscription_id']
 
-        if self.credentials.get('client_id') is not None and \
-           self.credentials.get('secret') is not None and \
-           self.credentials.get('tenant') is not None:
+        # get authentication authority
+        # for adfs, user could pass in authority or not.
+        # for others, use default authority from cloud environment
+        if self.credentials.get('adfs_authority_url'):
+            self._adfs_authority_url = self.credentials.get('adfs_authority_url')
+        else:
+            self._adfs_authority_url = self._cloud_environment.endpoints.active_directory
+
+        # get resource from cloud environment
+        self._resource = self._cloud_environment.endpoints.active_directory_resource_id
+
+        if self.credentials.get('credentials'):
+            self.azure_credentials = self.credentials.get('credentials')
+        elif self.credentials.get('client_id') and self.credentials.get('secret') and self.credentials.get('tenant'):
             self.azure_credentials = ServicePrincipalCredentials(client_id=self.credentials['client_id'],
                                                                  secret=self.credentials['secret'],
                                                                  tenant=self.credentials['tenant'],
                                                                  cloud_environment=self._cloud_environment)
+
+        elif self.credentials.get('ad_user') is not None and \
+                self.credentials.get('password') is not None and \
+                self.credentials.get('client_id') is not None and \
+                self.credentials.get('tenant') is not None:
+
+                self.azure_credentials = self.acquire_token_with_username_password(
+                    self._adfs_authority_url,
+                    self._resource,
+                    self.credentials['ad_user'],
+                    self.credentials['password'],
+                    self.credentials['client_id'],
+                    self.credentials['tenant'])
+
         elif self.credentials.get('ad_user') is not None and self.credentials.get('password') is not None:
             tenant = self.credentials.get('tenant')
             if not tenant:
@@ -315,9 +364,12 @@ class AzureRM(object):
                                                          self.credentials['password'],
                                                          tenant=tenant,
                                                          cloud_environment=self._cloud_environment)
+
         else:
             self.fail("Failed to authenticate with provided credentials. Some attributes were missing. "
-                      "Credentials must include client_id, secret and tenant or ad_user and password.")
+                      "Credentials must include client_id, secret and tenant or ad_user and password, or "
+                      "ad_user, password, client_id, tenant and adfs_authority_url(optional) for ADFS authentication, or "
+                      "be logged in using AzureCLI.")
 
     def log(self, msg):
         if self.debug:
@@ -361,6 +413,32 @@ class AzureRM(object):
 
         return None
 
+    def _get_azure_cli_credentials(self):
+        credentials, subscription_id = get_azure_cli_credentials()
+        cloud_environment = get_cli_active_cloud()
+
+        cli_credentials = {
+            'credentials': credentials,
+            'subscription_id': subscription_id,
+            'cloud_environment': cloud_environment
+        }
+        return cli_credentials
+
+    def _get_msi_credentials(self, subscription_id_param=None):
+        credentials = MSIAuthentication()
+        subscription_id_param = subscription_id_param or os.environ.get(AZURE_CREDENTIAL_ENV_MAPPING['subscription_id'], None)
+        try:
+            # try to get the subscription in MSI to test whether MSI is enabled
+            subscription_client = SubscriptionClient(credentials)
+            subscription = next(subscription_client.subscriptions.list())
+            subscription_id = str(subscription.subscription_id)
+            return {
+                'credentials': credentials,
+                'subscription_id': subscription_id_param or subscription_id
+            }
+        except Exception as exc:
+            return None
+
     def _get_credentials(self, params):
         # Get authentication credentials.
         # Precedence: cmd line parameters-> environment variables-> default profile in ~/.azure/credentials.
@@ -397,8 +475,31 @@ class AzureRM(object):
             self.log('Retrieved default profile credentials from ~/.azure/credentials.')
             return default_credentials
 
+        msi_credentials = self._get_msi_credentials(arg_credentials.get('subscription_id'))
+        if msi_credentials:
+            self.log('Retrieved credentials from MSI.')
+            return msi_credentials
+
+        try:
+            if HAS_AZURE_CLI_CORE:
+                self.log('Retrieving credentials from AzureCLI profile')
+            cli_credentials = self._get_azure_cli_credentials()
+            return cli_credentials
+        except CLIError as ce:
+            self.log('Error getting AzureCLI profile credentials - {0}'.format(ce))
+
         return None
 
+    def acquire_token_with_username_password(self, authority, resource, username, password, client_id, tenant):
+        authority_uri = authority
+
+        if tenant is not None:
+            authority_uri = authority + '/' + tenant
+
+        context = AuthenticationContext(authority_uri)
+        token_response = context.acquire_token_with_username_password(resource, username, password, client_id)
+        return AADTokenCredentials(token_response)
+
     def _register(self, key):
         try:
             # We have to perform the one-time registration here. Otherwise, we receive an error the first
@@ -412,16 +513,21 @@ class AzureRM(object):
                       "https://docs.microsoft.com/azure/azure-resource-manager/"
                       "resource-manager-common-deployment-errors#noregisteredproviderfound"))
 
+    def get_mgmt_svc_client(self, client_type, base_url, api_version):
+        client = client_type(self.azure_credentials,
+                             self.subscription_id,
+                             base_url=base_url,
+                             api_version=api_version)
+        client.config.add_user_agent(ANSIBLE_USER_AGENT)
+        return client
+
     @property
     def network_client(self):
         self.log('Getting network client')
         if not self._network_client:
-            self._network_client = NetworkManagementClient(
-                self.azure_credentials,
-                self.subscription_id,
-                base_url=self._cloud_environment.endpoints.resource_manager,
-                api_version='2017-06-01'
-            )
+            self._network_client = self.get_mgmt_svc_client(NetworkManagementClient,
+                                                            self._cloud_environment.endpoints.resource_manager,
+                                                            '2017-06-01')
             self._register('Microsoft.Network')
         return self._network_client
 
@@ -429,24 +535,18 @@ class AzureRM(object):
     def rm_client(self):
         self.log('Getting resource manager client')
         if not self._resource_client:
-            self._resource_client = ResourceManagementClient(
-                self.azure_credentials,
-                self.subscription_id,
-                base_url=self._cloud_environment.endpoints.resource_manager,
-                api_version='2017-05-10'
-            )
+            self._resource_client = self.get_mgmt_svc_client(ResourceManagementClient,
+                                                             self._cloud_environment.endpoints.resource_manager,
+                                                             '2017-05-10')
         return self._resource_client
 
     @property
     def compute_client(self):
         self.log('Getting compute client')
         if not self._compute_client:
-            self._compute_client = ComputeManagementClient(
-                self.azure_credentials,
-                self.subscription_id,
-                base_url=self._cloud_environment.endpoints.resource_manager,
-                api_version='2017-03-30'
-            )
+            self._compute_client = self.get_mgmt_svc_client(ComputeManagementClient,
+                                                            self._cloud_environment.endpoints.resource_manager,
+                                                            '2017-03-30')
             self._register('Microsoft.Compute')
         return self._compute_client
 
@@ -473,9 +573,11 @@ class AzureInventory(object):
         self.replace_dash_in_groups = False
         self.group_by_resource_group = True
         self.group_by_location = True
+        self.group_by_os_family = True
         self.group_by_security_group = True
         self.group_by_tag = True
         self.include_powerstate = True
+        self.use_private_ip = False
 
         self._inventory = dict(
             _meta=dict(
@@ -528,6 +630,8 @@ class AzureInventory(object):
                             help='Active Directory User')
         parser.add_argument('--password', action='store',
                             help='password')
+        parser.add_argument('--adfs_authority_url', action='store',
+                            help='Azure ADFS authority url')
         parser.add_argument('--cloud_environment', action='store',
                             help='Azure Cloud Environment name or metadata discovery URL')
         parser.add_argument('--resource-groups', action='store',
@@ -545,7 +649,7 @@ class AzureInventory(object):
             # get VMs for requested resource groups
             for resource_group in self.resource_groups:
                 try:
-                    virtual_machines = self._compute_client.virtual_machines.list(resource_group)
+                    virtual_machines = self._compute_client.virtual_machines.list(resource_group.lower())
                 except Exception as exc:
                     sys.exit("Error: fetching virtual machines for resource group {0} - {1}".format(resource_group, str(exc)))
                 if self._args.host or self.tags:
@@ -604,7 +708,7 @@ class AzureInventory(object):
 
             host_vars['os_disk'] = dict(
                 name=machine.storage_profile.os_disk.name,
-                operating_system_type=machine.storage_profile.os_disk.os_type.value
+                operating_system_type=machine.storage_profile.os_disk.os_type.value.lower()
             )
 
             if self.include_powerstate:
@@ -651,12 +755,15 @@ class AzureInventory(object):
                     for ip_config in network_interface.ip_configurations:
                         host_vars['private_ip'] = ip_config.private_ip_address
                         host_vars['private_ip_alloc_method'] = ip_config.private_ip_allocation_method
+                        if self.use_private_ip:
+                            host_vars['ansible_host'] = ip_config.private_ip_address
                         if ip_config.public_ip_address:
                             public_ip_reference = self._parse_ref_id(ip_config.public_ip_address.id)
                             public_ip_address = self._network_client.public_ip_addresses.get(
                                 public_ip_reference['resourceGroups'],
                                 public_ip_reference['publicIPAddresses'])
-                            host_vars['ansible_host'] = public_ip_address.ip_address
+                            if not self.use_private_ip:
+                                host_vars['ansible_host'] = public_ip_address.ip_address
                             host_vars['public_ip'] = public_ip_address.ip_address
                             host_vars['public_ip_name'] = public_ip_address.name
                             host_vars['public_ip_alloc_method'] = public_ip_address.public_ip_allocation_method
@@ -706,10 +813,16 @@ class AzureInventory(object):
 
         host_name = self._to_safe(vars['name'])
         resource_group = self._to_safe(vars['resource_group'])
+        operating_system_type = self._to_safe(vars['os_disk']['operating_system_type'].lower())
         security_group = None
         if vars.get('security_group'):
             security_group = self._to_safe(vars['security_group'])
 
+        if self.group_by_os_family:
+            if not self._inventory.get(operating_system_type):
+                self._inventory[operating_system_type] = []
+            self._inventory[operating_system_type].append(host_name)
+
         if self.group_by_resource_group:
             if not self._inventory.get(resource_group):
                 self._inventory[resource_group] = []

awx/plugins/library/scan_services.py

@@ -69,7 +69,7 @@ class ServiceScanService(BaseService):
 
         # Upstart
         if initctl_path is not None and chkconfig_path is None:
-            p = re.compile('^\s?(?P<name>.*)\s(?P<goal>\w+)\/(?P<state>\w+)(\,\sprocess\s(?P<pid>[0-9]+))?\s*$')
+            p = re.compile(r'^\s?(?P<name>.*)\s(?P<goal>\w+)\/(?P<state>\w+)(\,\sprocess\s(?P<pid>[0-9]+))?\s*$')
             rc, stdout, stderr = self.module.run_command("%s list" % initctl_path)
             real_stdout = stdout.replace("\r","")
             for line in real_stdout.split("\n"):
@@ -90,8 +90,8 @@ class ServiceScanService(BaseService):
         elif chkconfig_path is not None:
             #print '%s --status-all | grep -E "is (running|stopped)"' % service_path
             p = re.compile(
-                '(?P<service>.*?)\s+[0-9]:(?P<rl0>on|off)\s+[0-9]:(?P<rl1>on|off)\s+[0-9]:(?P<rl2>on|off)\s+'
-                '[0-9]:(?P<rl3>on|off)\s+[0-9]:(?P<rl4>on|off)\s+[0-9]:(?P<rl5>on|off)\s+[0-9]:(?P<rl6>on|off)')
+                r'(?P<service>.*?)\s+[0-9]:(?P<rl0>on|off)\s+[0-9]:(?P<rl1>on|off)\s+[0-9]:(?P<rl2>on|off)\s+'
+                r'[0-9]:(?P<rl3>on|off)\s+[0-9]:(?P<rl4>on|off)\s+[0-9]:(?P<rl5>on|off)\s+[0-9]:(?P<rl6>on|off)')
             rc, stdout, stderr = self.module.run_command('%s' % chkconfig_path, use_unsafe_shell=True)
             # Check for special cases where stdout does not fit pattern
             match_any = False
@@ -99,7 +99,7 @@ class ServiceScanService(BaseService):
                 if p.match(line):
                     match_any = True
             if not match_any:
-                p_simple = re.compile('(?P<service>.*?)\s+(?P<rl0>on|off)')
+                p_simple = re.compile(r'(?P<service>.*?)\s+(?P<rl0>on|off)')
                 match_any = False
                 for line in stdout.split('\n'):
                     if p_simple.match(line):

awx/settings/defaults.py

@@ -4,7 +4,6 @@
 import os
 import re  # noqa
 import sys
-import djcelery
 import six
 from datetime import timedelta
 
@@ -26,6 +25,8 @@ BASE_DIR = os.path.dirname(os.path.dirname(__file__))
 def is_testing(argv=None):
     import sys
     '''Return True if running django or py.test unit tests.'''
+    if 'PYTEST_CURRENT_TEST' in os.environ.keys():
+        return True
     argv = sys.argv if argv is None else argv
     if len(argv) >= 1 and ('py.test' in argv[0] or 'py/test.py' in argv[0]):
         return True
@@ -60,7 +61,7 @@ DATABASES = {
         'NAME': os.path.join(BASE_DIR, 'awx.sqlite3'),
         'ATOMIC_REQUESTS': True,
         'TEST': {
-            # Test database cannot be :memory: for celery/inventory tests.
+            # Test database cannot be :memory: for inventory tests.
             'NAME': os.path.join(BASE_DIR, 'awx_test.sqlite3'),
         },
     }
@@ -261,7 +262,6 @@ MIDDLEWARE_CLASSES = (  # NOQA
     'awx.sso.middleware.SocialAuthMiddleware',
     'crum.CurrentRequestUserMiddleware',
     'awx.main.middleware.URLModificationMiddleware',
-    'awx.main.middleware.DeprecatedAuthTokenMiddleware',
     'awx.main.middleware.SessionTimeoutMiddleware',
 )
 
@@ -280,7 +280,6 @@ INSTALLED_APPS = (
     'oauth2_provider',
     'rest_framework',
     'django_extensions',
-    'djcelery',
     'channels',
     'polymorphic',
     'taggit',
@@ -459,40 +458,9 @@ DEVSERVER_DEFAULT_PORT = '8013'
 # Set default ports for live server tests.
 os.environ.setdefault('DJANGO_LIVE_TEST_SERVER_ADDRESS', 'localhost:9013-9199')
 
-djcelery.setup_loader()
-
 BROKER_POOL_LIMIT = None
 BROKER_URL = 'amqp://guest:guest@localhost:5672//'
-CELERY_EVENT_QUEUE_TTL = 5
 CELERY_DEFAULT_QUEUE = 'awx_private_queue'
-CELERY_DEFAULT_EXCHANGE = 'awx_private_queue'
-CELERY_DEFAULT_ROUTING_KEY = 'awx_private_queue'
-CELERY_DEFAULT_EXCHANGE_TYPE = 'direct'
-CELERY_TASK_SERIALIZER = 'json'
-CELERY_RESULT_SERIALIZER = 'json'
-CELERY_ACCEPT_CONTENT = ['json']
-CELERY_TRACK_STARTED = True
-CELERYD_TASK_TIME_LIMIT = None
-CELERYD_TASK_SOFT_TIME_LIMIT = None
-CELERYD_POOL_RESTARTS = True
-CELERYD_AUTOSCALER = 'awx.main.utils.autoscale:DynamicAutoScaler'
-CELERY_RESULT_BACKEND = 'djcelery.backends.database:DatabaseBackend'
-CELERY_IMPORTS = ('awx.main.scheduler.tasks',)
-CELERY_QUEUES = ()
-CELERY_ROUTES = ('awx.main.utils.ha.AWXCeleryRouter',)
-
-
-def log_celery_failure(*args):
-    # Import annotations lazily to avoid polluting the `awx.settings` namespace
-    # and causing circular imports
-    from awx.main.tasks import log_celery_failure
-    return log_celery_failure(*args)
-
-
-CELERY_ANNOTATIONS = {'*': {'on_failure': log_celery_failure}}
-
-CELERYBEAT_SCHEDULER = 'celery.beat.PersistentScheduler'
-CELERYBEAT_MAX_LOOP_INTERVAL = 60
 CELERYBEAT_SCHEDULE = {
     'tower_scheduler': {
         'task': 'awx.main.tasks.awx_periodic_scheduler',
@@ -525,9 +493,6 @@ CELERYBEAT_SCHEDULE = {
 }
 AWX_INCONSISTENT_TASK_INTERVAL = 60 * 3
 
-# Celery queues that will always be listened to by celery workers
-# Note: Broadcast queues have unique, auto-generated names, with the alias
-# property value of the original queue name.
 AWX_CELERY_QUEUES_STATIC = [
     six.text_type(CELERY_DEFAULT_QUEUE),
 ]
@@ -626,8 +591,8 @@ SOCIAL_AUTH_SAML_ENABLED_IDPS = {}
 SOCIAL_AUTH_SAML_ORGANIZATION_ATTR = {}
 SOCIAL_AUTH_SAML_TEAM_ATTR = {}
 
-# Any ANSIBLE_* settings will be passed to the subprocess environment by the
-# celery task.
+# Any ANSIBLE_* settings will be passed to the task runner subprocess
+# environment
 
 # Do not want AWX to ask interactive questions and want it to be friendly with
 # reprovisioning
@@ -641,8 +606,7 @@ ANSIBLE_PARAMIKO_RECORD_HOST_KEYS = False
 # output
 ANSIBLE_FORCE_COLOR = True
 
-# Additional environment variables to be passed to the subprocess started by
-# the celery task.
+# Additional environment variables to be passed to the ansible subprocesses
 AWX_TASK_ENV = {}
 
 # Flag to enable/disable updating hosts M2M when saving job events.
@@ -1027,7 +991,10 @@ LOGGING = {
         'timed_import': {
             '()': 'awx.main.utils.formatters.TimeFormatter',
             'format': '%(relativeSeconds)9.3f %(levelname)-8s %(message)s'
-        }
+        },
+        'dispatcher': {
+            'format': '%(asctime)s %(levelname)-8s %(name)s PID:%(process)d %(message)s',
+        },
     },
     'handlers': {
         'console': {
@@ -1077,6 +1044,19 @@ LOGGING = {
             'backupCount': 5,
             'formatter':'simple',
         },
+        'dispatcher': {
+            'level': 'WARNING',
+            'class':'logging.handlers.RotatingFileHandler',
+            'filters': ['require_debug_false'],
+            'filename': os.path.join(LOG_ROOT, 'dispatcher.log'),
+            'maxBytes': 1024 * 1024 * 5, # 5 MB
+            'backupCount': 5,
+            'formatter':'dispatcher',
+        },
+        'celery.beat': {
+            'class':'logging.StreamHandler',
+            'level': 'ERROR'
+        },  # don't log every celerybeat wakeup
         'inventory_import': {
             'level': 'DEBUG',
             'class':'logging.StreamHandler',
@@ -1158,8 +1138,13 @@ LOGGING = {
         },
         'awx.main': {
             'handlers': ['null']
-        }, 'awx.main.commands.run_callback_receiver': {
+        },
+        'awx.main.commands.run_callback_receiver': {
             'handlers': ['callback_receiver'],
+            'level': 'INFO'  # in debug mode, includes full callback data
+        },
+        'awx.main.dispatch': {
+            'handlers': ['dispatcher'],
         },
         'awx.isolated.manager.playbooks': {
             'handlers': ['management_playbooks'],

awx/settings/development.py

@@ -68,13 +68,6 @@ template['OPTIONS']['loaders'] = (
     'django.template.loaders.app_directories.Loader',
 )
 
-# Disable capturing all SQL queries when running celeryd in development.
-if 'celery' in sys.argv:
-    SQL_DEBUG = False
-
-CELERYD_HIJACK_ROOT_LOGGER = False
-CELERYD_LOG_COLOR = True
-
 CALLBACK_QUEUE = "callback_tasks"
 
 # Enable dynamically pulling roles from a requirement.yml file
@@ -149,15 +142,6 @@ except ImportError:
 
 CLUSTER_HOST_ID = socket.gethostname()
 
-# Supervisor service name dictionary used for programatic restart
-SERVICE_NAME_DICT = {
-    "celery": "celery",
-    "callback": "receiver",
-    "runworker": "channels",
-    "uwsgi": "uwsgi",
-    "daphne": "daphne",
-    "nginx": "nginx"}
-
 try:
     socket.gethostbyname('docker.for.mac.host.internal')
     os.environ['SDB_NOTIFY_HOST'] = 'docker.for.mac.host.internal'

awx/settings/local_settings.py.docker_compose

@@ -73,13 +73,13 @@ if "pytest" in sys.modules:
             'ENGINE': 'django.db.backends.sqlite3',
             'NAME': os.path.join(BASE_DIR, 'awx.sqlite3'),
             'TEST': {
-                # Test database cannot be :memory: for celery/inventory tests.
+                # Test database cannot be :memory: for inventory tests.
                 'NAME': os.path.join(BASE_DIR, 'awx_test.sqlite3'),
             },
         }
     }
 
-# Celery AMQP configuration.
+# AMQP configuration.
 BROKER_URL = "amqp://{}:{}@{}/{}".format(os.environ.get("RABBITMQ_USER"),
                                   os.environ.get("RABBITMQ_PASS"),
                                   os.environ.get("RABBITMQ_HOST"),
@@ -138,8 +138,7 @@ REMOTE_HOST_HEADERS = ['REMOTE_ADDR', 'REMOTE_HOST']
 # REMOTE_HOST_HEADERS will be trusted unconditionally')
 PROXY_IP_WHITELIST = []
 
-# Define additional environment variables to be passed to subprocess started by
-# the celery task.
+# Define additional environment variables to be passed to ansible subprocesses
 #AWX_TASK_ENV['FOO'] = 'BAR'
 
 # If set, use -vvv for project updates instead of -v for more output.
@@ -197,7 +196,7 @@ LOGGING['handlers']['syslog'] = {
 LOGGING['loggers']['django.request']['handlers'] = ['console']
 LOGGING['loggers']['rest_framework.request']['handlers'] = ['console']
 LOGGING['loggers']['awx']['handlers'] = ['console', 'external_logger']
-LOGGING['loggers']['awx.main.commands.run_callback_receiver']['handlers'] = ['console']
+LOGGING['loggers']['awx.main.commands.run_callback_receiver']['handlers'] = []  # propogates to awx
 LOGGING['loggers']['awx.main.tasks']['handlers'] = ['console', 'external_logger']
 LOGGING['loggers']['awx.main.scheduler']['handlers'] = ['console', 'external_logger']
 LOGGING['loggers']['django_auth_ldap']['handlers'] = ['console']

awx/settings/local_settings.py.example

@@ -39,13 +39,13 @@ if is_testing(sys.argv):
             'ENGINE': 'django.db.backends.sqlite3',
             'NAME': os.path.join(BASE_DIR, 'awx.sqlite3'),
             'TEST': {
-                # Test database cannot be :memory: for celery/inventory tests.
+                # Test database cannot be :memory: for tests.
                 'NAME': os.path.join(BASE_DIR, 'awx_test.sqlite3'),
             },
         }
     }
     
-# Celery AMQP configuration.
+# AMQP configuration.
 BROKER_URL = 'amqp://guest:guest@localhost:5672'
 
 # Set True to enable additional logging from the job_event_callback plugin
@@ -94,8 +94,7 @@ REMOTE_HOST_HEADERS = ['REMOTE_ADDR', 'REMOTE_HOST']
 # REMOTE_HOST_HEADERS will be trusted unconditionally')
 PROXY_IP_WHITELIST = []
 
-# Define additional environment variables to be passed to subprocess started by
-# the celery task.
+# Define additional environment variables to be passed to ansible subprocesses
 #AWX_TASK_ENV['FOO'] = 'BAR'
 
 # If set, use -vvv for project updates instead of -v for more output.

awx/settings/production.py

@@ -54,21 +54,13 @@ AWX_ISOLATED_USERNAME = 'awx'
 
 LOGGING['handlers']['tower_warnings']['filename'] = '/var/log/tower/tower.log'
 LOGGING['handlers']['callback_receiver']['filename'] = '/var/log/tower/callback_receiver.log'
+LOGGING['handlers']['dispatcher']['filename'] = '/var/log/tower/dispatcher.log'
 LOGGING['handlers']['task_system']['filename'] = '/var/log/tower/task_system.log'
 LOGGING['handlers']['fact_receiver']['filename'] = '/var/log/tower/fact_receiver.log'
 LOGGING['handlers']['management_playbooks']['filename'] = '/var/log/tower/management_playbooks.log'
 LOGGING['handlers']['system_tracking_migrations']['filename'] = '/var/log/tower/tower_system_tracking_migrations.log'
 LOGGING['handlers']['rbac_migrations']['filename'] = '/var/log/tower/tower_rbac_migrations.log'
 
-# Supervisor service name dictionary used for programatic restart
-SERVICE_NAME_DICT = {
-    "beat": "awx-celery-beat",
-    "celery": "awx-celery",
-    "callback": "awx-callback-receiver",
-    "channels": "awx-channels-worker",
-    "uwsgi": "awx-uwsgi",
-    "daphne": "awx-daphne"}
-
 # Store a snapshot of default settings at this point before loading any
 # customizable config files.
 DEFAULTS_SNAPSHOT = {}

awx/sso/ldap_group_types.py

@@ -42,7 +42,7 @@ class PosixUIDGroupType(LDAPGroupType):
                 )
 
             search = group_search.search_with_additional_term_string(filterstr)
-            search.attrlist = [self.name_attr]
+            search.attrlist = [str(self.name_attr)]
             groups = search.execute(ldap_user.connection)
         except (KeyError, IndexError):
             pass

awx/ui/README.md

@@ -1,46 +1,26 @@
 # AWX UI
 
 ## Requirements
+- node.js 8.x LTS
+- npm 5.x LTS
+- bzip2, gcc-c++, git, make
 
-### Node / NPM
+## Development
+The API development server will need to be running. See [CONTRIBUTING.md](../../CONTRIBUTING.md).
 
-AWX currently requires the 6.x LTS version of Node and NPM.
+```shell
+# Build ui for the devel environment - reachable at https://localhost:8043
+make ui-devel
 
-macOS installer: [https://nodejs.org/dist/latest-v6.x/](https://nodejs.org/dist/latest-v6.x/)
+# Alternatively, start the ui development server. While running, the ui will be reachable
+# at https://localhost:3000 and updated automatically when code changes.
+make ui-docker
 
-RHEL / CentOS / Fedora:
-
-```
-$ curl --silent --location https://rpm.nodesource.com/setup_6.x | bash -
-$ yum install nodejs
-```
-
-### Other Dependencies
-
-On macOS, install the Command Line Tools:
-
-```
-$ xcode-select --install
-```
-
-RHEL / CentOS / Fedora:
-
-```
-$ yum install bzip2 gcc-c++ git make
-```
-
-## Usage
-
-### Starting the UI
-
-First, the AWX API will need to be running. See [CONTRIBUTING.md](../../CONTRIBUTING.md).
-
-When using Docker for Mac or native Docker on Linux:
-
-```
-$ make ui-docker
+# When using docker machine, use this command to start the ui development server instead.
+DOCKER_MACHINE_NAME=default make ui-docker-machine
 ```
 
+## Development with an external server
 If you normally run awx on an external host/server (in this example, `awx.local`),
 you'll need to reconfigure the webpack proxy slightly for `make ui-docker` to
 work:
@@ -63,119 +43,32 @@ awx/ui/build/webpack.watch.js
    },
 ```
 
-When using Docker Machine:
+## Testing
+```shell
+# run linters
+make jshint
 
-```
-$ DOCKER_MACHINE_NAME=default make ui-docker-machine
+# run unit tests
+make ui-test-ci
+
+# run e2e tests - see awx/ui/test/e2e for more information
+npm --prefix awx/ui run e2e
 ```
 
-### Running Tests
+## Adding dependencies
+```shell
+# add a development or build dependency
+npm install --prefix awx/ui --save-dev dev-package@1.2.3
 
-Run unit tests locally, poll for changes to both source and test files, launch tests in supported browser engines:
+# add a production dependency
+npm install --prefix awx/ui --save prod-package@1.23
 
-```
-$ make ui-test
+# add the updated package.json and lock files to scm
+git add awx/ui/package.json awx/ui/package-lock.json
 ```
 
-Run unit tests in a CI environment (Jenkins)
-
+## Building for Production
+```shell
+# built files are placed in awx/ui/static
+make ui-release
 ```
-$ make ui-test-ci
-```
-
-### Adding new dependencies
-
-
-#### Add / update a bundled vendor dependency
-
-1. `npm install --prefix awx/ui --save some-frontend-package@1.2.3`
-2. Add `'some-package'` to `var vendorFiles` in `./grunt-tasks/webpack.js`
-3. `npm  --prefix awx/ui shrinkwrap` to freeze current dependency resolution
-
-#### Add / update a dependecy in the build/test pipeline
-
-1. `npm install --prefix awx/ui --save-dev some-toolchain-package@1.2.3`
-2. `npm  --prefix awx/ui shrinkwrap` to freeze current dependency resolution
-
-### Polyfills, shims, patches
-
-The Webpack pipeline will prefer module patterns in this order: CommonJS, AMD, UMD. For a comparison of supported patterns, refer to [https://webpack.github.io/docs/comparison.html](Webpack's docs).
-
-Some javascript libraries do not export their contents as a module, or depend on other third-party components. If the library maintainer does not wrap their lib in a factory that provides a CommonJS or AMD module, you will need to provide dependencies with a shim.
-
-1. Shim implicit dependencies using Webpack's [ProvidePlugin](https://github.com/webpack/webpack/blob/006d59500de0493c4096d5d4cecd64eb12db2b95/lib/ProvidePlugin.js). Example:
-
-```js
-// AWX source code depends on the lodash library being available as _
-_.uniq([1,2,3,1]) // will throw error undefined
-```
-
-```js
-// webpack.config.js
-plugins: [
-  new webpack.ProvidePlugin({
-      '_': 'lodash',
-  })
-]
-```
-
-```js
-// the following requirement is inserted by webpack at build time
-var _ = require('lodash');
-_.uniq([1,2,3,1])
-```
-
-2. Use [`imports-loader`](https://webpack.github.io/docs/shimming-modules.html#importing) to inject requirements into the namespace of vendor code at import time. Use [`exports-loader`](https://webpack.github.io/docs/shimming-modules.html#exporting) to conventionally export vendor code lacking a conventional export pattern.
-3. [Apply a functional patch](https://gist.github.com/leigh-johnson/070159d3fd780d6d8da6e13625234bb3). A webpack plugin is the correct choice for a functional patch if your patch needs to access events in a build's lifecycle. A webpack loader is preferable if you need to compile and export a custom pattern of library modules.
-4. [Submit patches to libraries without modular exports](https://github.com/leigh-johnson/ngToast/commit/fea95bb34d27687e414619b4f72c11735d909f93) - the internet will thank you
-
-Some javascript libraries might only get one module pattern right.
-
-### Environment configuration - used in development / test builds
-
-Build tasks are parameterized with environment variables.
-
-`package.json` contains default environment configuration. When `npm run myScriptName` is executed, these variables will be exported to your environment with the prefix `npm_package_config_`. For example, `my_variable` will be exported to `npm_package_config_my_variable`.
-
-Environment variables can accessed in a Javascript via `PROCESS.env`.
-
-``` json
-  "config": {
-    "django_port": "8013",
-    "websocket_port": "8080",
-    "django_host": "0.0.0.0"
-  }
-```
-
-Example usage in `npm run build-docker-machine`:
-
-```bash
-$ docker-machine ssh $DOCKER_MACHINE_NAME -f -N -L ${npm_package_config_websocket_port}:localhost:${npm_package_config_websocket_port}; ip=$(docker-machine ip $DOCKER_MACHINE_NAME); echo npm set awx:django_host ${ip}; $ grunt dev
-```
-
-Example usage in an `npm test` script target:
-
-```
-npm_package_config_websocket_port=mock_websocket_port npm_package_config_django_port=mock_api_port npm_package_config_django_host=mock_api_host npm run test:someMockIntegration
-```
-
-You'll usually want to pipe and set vars prior to running a script target:
-```
-$ npm set awx:websocket_host ${mock_host}; npm run script-name
-```
-
-### NPM Scripts
-
-Examples:
-```json
-  {
-    "scripts": {
-      "pretest": "echo I run immediately before 'npm test' executes",
-      "posttest": "echo I run immediately after 'npm test' exits",
-      "test": "karma start karma.conf.js"
-    }
-  }
-```
-
-`npm test` is an alias for `npm run test`. Refer to [script field docs](https://docs.npmjs.com/misc/scripts) for a list of other runtime events.
-

awx/ui/client/features/index.js

@@ -9,6 +9,7 @@ import atFeaturesTemplates from '~features/templates';
 import atFeaturesUsers from '~features/users';
 import atFeaturesJobs from '~features/jobs';
 import atFeaturesPortalMode from '~features/portalMode';
+import atFeaturesProjects from '~features/projects';
 
 const MODULE_NAME = 'at.features';
 
@@ -24,6 +25,7 @@ angular.module(MODULE_NAME, [
     atFeaturesOutput,
     atFeaturesTemplates,
     atFeaturesPortalMode,
+    atFeaturesProjects
 ]);
 
 export default MODULE_NAME;

awx/ui/client/features/jobs/index.js

@@ -1,12 +1,14 @@
 import JobsStrings from './jobs.strings';
 import jobsRoute from './routes/jobs.route';
 import { jobsSchedulesRoute, jobsSchedulesEditRoute } from '../../src/scheduler/schedules.route';
+import jobsListController from './jobsList.controller';
 
 const MODULE_NAME = 'at.features.jobs';
 
 angular
     .module(MODULE_NAME, [])
     .service('JobsStrings', JobsStrings)
+    .controller('jobsListController', jobsListController)
     .run(['$stateExtender', ($stateExtender) => {
         $stateExtender.addState(jobsRoute);
         $stateExtender.addState(jobsSchedulesRoute);

awx/ui/client/features/jobs/jobsList.controller.js

@@ -76,6 +76,22 @@ function ListJobsController (
             return { icon, link, value };
         });
 
+    vm.getSliceJobDetails = (job) => {
+        if (!job.job_slice_count) {
+            return null;
+        }
+
+        if (job.job_slice_count === 1) {
+            return null;
+        }
+
+        if (job.job_slice_number && job.job_slice_count) {
+            return `Slice Job ${job.job_slice_number}/${job.job_slice_count}`;
+        }
+
+        return null;
+    };
+
     vm.getSref = ({ type, id }) => {
         let sref;
 

awx/ui/client/features/jobs/jobsList.view.html

@@ -23,7 +23,8 @@
                     status-tip="{{ vm.strings.get('list.STATUS_TOOLTIP', job.status) }}"
                     header-value="{{ job.id }} - {{ job.name }}"
                     header-state="{{ vm.getSref(job) }}"
-                    header-tag="{{ vm.jobTypes[job.type] }}">
+                    header-tag="{{ vm.jobTypes[job.type] }}"
+                    secondary-tag="{{ vm.getSliceJobDetails(job) }}">
                 </at-row-item>
                 <div class="at-Row--inline">
                     <at-row-item

awx/ui/client/features/jobs/routes/jobs.route.js

@@ -1,5 +1,4 @@
 import { N_ } from '../../../src/i18n';
-import jobsListController from '../jobsList.controller';
 import indexController from '../index.controller';
 
 const indexTemplate = require('~features/jobs/index.view.html');
@@ -69,7 +68,7 @@ export default {
         },
         'jobsList@jobs': {
             templateUrl: jobsListTemplate,
-            controller: jobsListController,
+            controller: 'jobsListController',
             controllerAs: 'vm'
         }
     }

awx/ui/client/features/output/details.component.js

@@ -126,6 +126,33 @@ function getSourceWorkflowJobDetails () {
     return { link, tooltip };
 }
 
+function getSliceJobDetails () {
+    const count = resource.model.get('job_slice_count');
+
+    if (!count) {
+        return null;
+    }
+
+    if (count === 1) {
+        return null;
+    }
+
+    const number = resource.model.get('job_slice_number');
+
+    if (!number) {
+        return null;
+    }
+
+    const label = strings.get('labels.SLICE_JOB');
+    const offset = `${number}/${count}`;
+    const tooltip = strings.get('tooltips.SLICE_JOB_DETAILS');
+
+    if (label && offset && tooltip) {
+        return { label, offset, tooltip };
+    }
+    return null;
+}
+
 function getJobTemplateDetails () {
     const jobTemplate = resource.model.get('summary_fields.job_template');
 
@@ -671,6 +698,7 @@ function JobDetailsController (
         vm.jobType = getJobTypeDetails();
         vm.jobTemplate = getJobTemplateDetails();
         vm.sourceWorkflowJob = getSourceWorkflowJobDetails();
+        vm.sliceJobDetails = getSliceJobDetails();
         vm.inventory = getInventoryDetails();
         vm.project = getProjectDetails();
         vm.projectUpdate = getProjectUpdateDetails();

awx/ui/client/features/output/details.partial.html

@@ -151,6 +151,12 @@
     <div class="JobResults-resultRowText">{{ vm.jobType.value }}</div>
 </div>
 
+<!-- SLICE JOB DETAIL -->
+<div class="JobResults-resultRow" ng-if="vm.sliceJobDetails">
+    <label class="JobResults-resultRowLabel">{{ vm.sliceJobDetails.label }}</label>
+    <div class="JobResults-resultRowText">{{ vm.sliceJobDetails.offset }}</div>
+</div>
+
 <!-- LAUNCHED BY DETAIL -->
 <div class="JobResults-resultRow" ng-if="vm.launchedBy">
     <label class="JobResults-resultRowLabel">{{ vm.launchedBy.label }}</label>

awx/ui/client/features/output/output.strings.js

@@ -23,6 +23,7 @@ function OutputStrings (BaseString) {
         EXTRA_VARS: t.s('Read-only view of extra variables added to the job template'),
         INVENTORY: t.s('View the Inventory'),
         JOB_TEMPLATE: t.s('View the Job Template'),
+        SLICE_JOB_DETAILS: t.s('Job is one of several from a JT that slices on inventory'),
         PROJECT: t.s('View the Project'),
         PROJECT_UPDATE: t.s('View Project checkout results'),
         SCHEDULE: t.s('View the Schedule'),
@@ -55,6 +56,7 @@ function OutputStrings (BaseString) {
         JOB_EXPLANATION: t.s('Explanation'),
         JOB_TAGS: t.s('Job Tags'),
         JOB_TEMPLATE: t.s('Job Template'),
+        SLICE_JOB: t.s('Slice Job'),
         JOB_TYPE: t.s('Job Type'),
         LABELS: t.s('Labels'),
         LAUNCHED_BY: t.s('Launched By'),

awx/ui/client/features/projects/index.controller.js

@@ -0,0 +1,19 @@
+function IndexProjectsController ($scope, strings, dataset) {
+    const vm = this;
+    vm.strings = strings;
+    vm.count = dataset.data.count;
+
+    $scope.$on('updateCount', (e, count) => {
+        if (typeof count === 'number') {
+            vm.count = count;
+        }
+    });
+}
+
+IndexProjectsController.$inject = [
+    '$scope',
+    'ProjectsStrings',
+    'Dataset',
+];
+
+export default IndexProjectsController;

awx/ui/client/features/projects/index.js

@@ -0,0 +1,9 @@
+import ProjectsStrings from './projects.strings';
+
+const MODULE_NAME = 'at.features.projects';
+
+angular
+    .module(MODULE_NAME, [])
+    .service('ProjectsStrings', ProjectsStrings);
+
+export default MODULE_NAME;