* Tower now injects facts into jobs via memcached for use by Ansible
playbooks. On the Ansible side, this is accomplished by the existing
mechanism, an Ansible Fact Cache Plugin + memcached. On the Tower side,
memcached is leveraged heavily.
* From an insights cred, access all related inventories via
cred.insights_inventories rather than cred.insights_credential
* This fixes auto-attaching insights_credential to credential summary
fields. Fixes by NOT attaching them. That is the desired behavior.
* During insights project updates, if the etag from insights matches the
last etag (scm_revision) then don't get all the maintenance playbooks.
Otherwise, get all the maintenance playbooks and update scm_revision
with the etag.
* A credential is considered an insights credential if a project of
scm_type 'insights' is associated.
* all non-insights Projects that point at an insights credential are
nulled
* Issue request to Red Hat Insights API from Tower
/hosts/<id>/insights endpoint. User the first found Satellite 6
Credential as Basic Auth requests to Insights API.
* We've switched from Q() to Host.objects.filter() and you can't do
~(Host.objects.filter()). We'd need to create an AST to track the
operations in order to perform the boolean logic ~ and we don't have
time for that now.
* Prefetch all Jobs Types related instance group
* Prefetch inventory updates inventory source. The attribute
inventory_source.inventory_id is accessed when building the hash
tables of the running tasks.
* When generating dependencies (i.e. dynamically launching Project
Update and Inventory Update) only create the dynamic dependencies if
update_on_launch is True.
* filter_host_string now returns a query set instead of Q(). This change
updates the Host manager to remove the .filter() call since the results
returned from filter_host_string() has already turned Q() -> Query Set
* Before, the boolean logic operators were performed against the Q()
objects, iteratively. Now, boolean logic is done after
Host.objects.filter(Q()). This results in the wanted and expected
results.
* Add inisghts fact scan module
* Update fact scan playbook to call new insight fact scan module
* JT run w/ store_facts=True will save scanned facts. We "skim" the
machine_id fact from Insights fact scans and associate it with the host
that the fact scan came from.
* Remove system tracking piggybacked most-recent-fact-gathering
* Add explicit per-jobtemplate configurable fact gathering
* Continue to support system tracking
* Fixed a bug where we weren't ack()ing system tracking rabbit msgs
* We create <project_name>.lock file in the tower project root
directory. This should not require that the project directory exists,
only that the project directory name is generatable.
* This returns /api/vx/hosts/ to having all the nice response items like
count, results, next, prev. I had "chopped" that off by explicitly only
returning result.
* Fixed a bug in the fact search language parser where it would not loop
over all the <and, or> operations. The result, we can support (a=b or
a=c or a=d) also support (a=b and b=b and c=b)
* jsonbfield supports json querying. jsonfield package is still a jsonb
postgres data type, but doesn't support jsonb style querying.
* add undo migration support to GIN index
* Copy of the most recent system tracking fact for each module type.
Utimately, this allows us to GIN index the jsonb object to support
fact searching.
* Generating the set of valid unicode characters is expensive in terms
of memory storage. Instead, we define the grammer by the negation of
allowed unicode characters. Much faster.
* Establish a base grammar for handling json path specification and
value matching. With boolean logic support and parenthesis grouping
i.e. (a.b.c="value") and ((a.b="foo") or (a="bar"))
* generate Q() results for passing to Host.objects.filter()
* Hooked up via /api/v1/hosts?host_filter=...
* DynamicFilterField added to store host filter as string with grammar
parser attached as static methods for later use by DynamicInventory &
continued use by host_filter=...
* Inventory updates that are canceled and are dependencies of jobs
result in the dependent job being canceled. This code adds a better
description to job_explanation so that a job marked as canceled can be
traced back to the inventory update that triggered this case.
* I explored the |urlencode() filter instead. Ansible and git didn't
like this.
* Quoting doesn't seemed to be required. Ansible and the git module are
smart enough to do the right thing. I tested with a space in the repo
url path and it works.
* We should not modify the original env because it's what is going to be
passed to the Ansible process. This will contain all of the passwords,
unobscured. The safe env is what will get saved into the model.
1. If the setting isn't in the database, then the specified default is
returned.
2. If the setting IS in the database, then the 'value' field, from the
database, is returned; Unless, the value is null in the database, then
the defined default is returned.
* Take case 1. If the default=None then an AttributeError is raised
* Thus, by setting default='', in case 1. We don't get an AttributeError
because '' will be returned.
* Also note that we don't allow_null=True because then the database
entry could be None. Accessing the setting would raise an AttributeError
in this case. Further, since we set default='' it doesn't matter if null
is in the database.
TL;DR set default=<something_other_than_None>
* Associate the celery_id with the job at the earliest point possible.
This ensures that a waiting job has a celery id. Thus, we are free to
fail waiting jobs that don't have a celery id.
Chris Meyers