For ad-hoc jobs where we are experiencing issues related to disabling
control persist to support proot. See commit
f69846e7871cae41b19ecda33563d6365dc9c3e0 for earlier analysis
If the celerybeat database becomes corrupt then it can prevent starting
some services for Tower. This happens in an obscure way and currently
that database is stored at the top of the source tree. This moves it
into the container managed space so that failure scenario won't persist
if the container is re-created
Somewhere recently we broke the ability to gather multiple modules in
one scan job pass. This fixes that, it also further improves the scan
job performance.
We now need to deal with two azure inventory scripts. The logic here
was incorrect after the PR importing the new module. This brings it
back in line with the old logic for selecting the class azure inventory script.
This deep copy would cause ansible-playbook to get hung up and fail to
spawn new subprocesses thus forcing extremely large scan jobs to
serialize their connections through one process. This had the effect
of forcing `forks=1` behavior.
This change was for citco, a customer who is running scan jobs against
600 hosts with a forks value of 200.
* release_2.4.5:
Remove distribute from the setup virtualenv if installed (#1631)
Update changelogs for 2.4.5 release
Only export changed targets in reprepro
Improve the efficiency of the stdout dump database migration
Added logic to not show the loop summary events in the UI by looking at event_data.event_loop.
Handle runner items from ansible v2
bump boto
fix case of Ansible v2 _result.cmd is list for release 2.4.5
Handle both string and list hosts in our hosts->name conversion
Default play names to the hosts the play was run against
Use better isinstance(x) type checking
Backporting test fixes from PR #1020: Fix error with ad hoc command events when running in check mode.
Fix for tasks breaking when using 'yum' with ansible 1.9.4
Bump 2.4.5 version, changelogs, and reprepo
* release_2.4.5:
Remove distribute from the setup virtualenv if installed (#1631)
Update changelogs for 2.4.5 release
Only export changed targets in reprepro
Improve the efficiency of the stdout dump database migration
Added logic to not show the loop summary events in the UI by looking at event_data.event_loop.
Handle runner items from ansible v2
bump boto
fix case of Ansible v2 _result.cmd is list for release 2.4.5
Handle both string and list hosts in our hosts->name conversion
Default play names to the hosts the play was run against
Use better isinstance(x) type checking
Backporting test fixes from PR #1020: Fix error with ad hoc command events when running in check mode.
Fix for tasks breaking when using 'yum' with ansible 1.9.4
Bump 2.4.5 version, changelogs, and reprepo
* Vendor ansible's azure_rm inventory script
* Add new inventory type
* Add new credential type
* Expand host instance_id column from varchar 100 to 1024 to accept the
long instance ids returned by Azure
* Make the inventory_import azure match rename more explicit.
Previously a job template would always block another job template launch
regardless of the details of the job template. We now restrict that
blocking logic to only block in the case that the job template was
launched with the same inventory. We keep the exclusion where if the
launch type is 'callback' and the limits differ then they won't be blocked
* Break requirements down into ansible and tower reqs
* Generate separate Ansible and Tower virtual environments
* Install appropriate requirements files into each one
* Modify development tools to use these venvs instead of our old
site-packages
* Modify settings to indicate venv enablement and location of venvs
* Modify tasks to use the proper virtual environment for its purpose
* Break requirements down into ansible and tower reqs
* Generate separate Ansible and Tower virtual environments
* Install appropriate requirements files into each one
* Modify development tools to use these venvs instead of our old
site-packages
* Modify settings to indicate venv enablement and location of venvs
* Modify tasks to use the proper virtual environment for its purpose
DRF will try to resolve potential candidates into the OPTIONS endpoint.
This is mainly to support their POST field in the browseable API. We
don't need this and it can yield some expensive queries so we bypass
generating choices for any RelatedField fields
For new installs the database may not be online and thus the migration
can't happen. This will allow it to continue but we may want more
messaging to the user running potential upgrades
Previously this would only be visible if survey is enabled. It should
be possible to define and view a survey even if this is disabled. The
enablement flag controls whether the survey prompt is shown and required.
* Break requirements down into ansible and tower reqs
* Generate separate Ansible and Tower virtual environments
* Install appropriate requirements files into each one
* Modify development tools to use these venvs instead of our old
site-packages
* Modify settings to indicate venv enablement and location of venvs
* Modify tasks to use the proper virtual environment for its purpose
* notifications_work: (23 commits)
Updates to notification unit tests after @wwitzel3's feedback
Fix some notifications issues and write some tests
Add notification system documentation
Clean up flake8 related issues
Fixing up some unicode issues
Implement tower ui view url on models
Sanity check and force proper types in admin check
Proper type for in check
Adding migration and base notification type
Add a periodic administrative notification
Refactor message generator
Support notification password field encryption
Notification configuration type checking
Refactor NotificationTemplate to Notifier
Implement irc notification backend
Add webhook notification backend
Pagerduty and Hipchat backends plus some cleanup
Notification serializers, views, and tasks
Implement notification serializer and validations
Notification endpoints and url expositions
...
Matthew Jones