| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently, docker_image_availability health_check
does not support authenticated registries.
This commit adds the '--creds=' option to skopeo
if needed to support authentication credentials.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1316341
Some other fixes to handle docker config better:
Should now account properly for blocked registries, insecure registries,
multiple additional registries, and oreg_url registry with or without credentials.
Output on failure should be clearer about what was tried.
Fixed a bug in the action_plugin_test exposed by these changes.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some refactoring of checks and the action plugin to enable writing files
locally about the check operation and results, if the user wants them.
This is aimed at enabling persistent and machine-readable results from
recurring runs of health checks.
Now, rather than trying to build a result hash to return from running
each check, checks can just register what they need to as they're going
along, and the action plugin processes state when the check is done.
Checks can register failures, notes about what they saw, and arbitrary
files to be saved into a directory structure where the user specifies.
If no directory is specified, no files are written.
At this time checks can still return a result hash, but that will likely
be refactored away in the next iteration.
Multiple failures can be registered without halting check execution.
Throwing an exception or returning a hash with "failed" is registered as
a failure.
execute_module now does a little more with the results. Results are
automatically included in notes and written individually as files.
"changed" results are propagated. Some json results are decoded.
A few of the checks were enhanced to use these features; all get some of
the features for free.
|
|
|
|
|
|
|
|
| |
Can now set openshift_disable_check=* to disable all checks without
needing to know their names.
fixes bug 1462106
https://bugzilla.redhat.com/show_bug.cgi?id=1462106
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This is a simple mechanism to learn what health checks are available.
Note that we defer task_vars verification, so that we can compute
requested_checks and resolved_checks earlier, allowing us to list checks
even if openshift_facts has not run.
|
|
|
|
|
|
|
|
|
|
| |
Introduced the 'changed' property for checks that can make changes to
track whether they did or not. Rather than the check's own logic having
to track this and include it in the result hash, just set the property
and have the action plugin insert it in the result hash after running
(even if there is an exception).
Cleared out a lot of crufty "changed: false" hash entries.
|
|
|
|
|
|
|
|
|
| |
Move task_vars into instance variable so we don't have to pass it
around everywhere. Also store tmp. Make sure both are filled in on
execute_module.
In the process, is_active became an instance method, and task_vars is
basically never used directly outside of test code.
|
|
|
|
|
|
|
| |
Calling the action plugin (e.g. when running a playbook) with an
incorrect check name was raising an unhandled exception, leading to poor
output in Ansible (requiring a higher verbosity level to see what is
going wrong).
|
|
|
|
|
| |
Added indicator to check result for why that check was skipped.
Note that currently the user will only see it with ansible-playbook -vv
|
|
|
|
|
|
|
|
| |
In Ansible 2.3+, the base action plugin class' run method depends
on accessing attributes (check_mode) of its play_context.
In older versions play_context was not involved in run, and thus None
was passed in.
|
| |
|
|
So that all names it return are directly usable, without checking for
existence.
|