Removing duplicate .txt extension from errors.txt..
Also, Adding errors.txt in artcl_logstash_files so that logstash pick
As size of logstash file is a factor, Adding errors.txt instead of
individual files seems better approach as errors.txt only contains
error/traceback from different log files.
We need to be able to see which jobs are failing
on tempest via tripleo-health. The initial
query would be the sova query ( ... FAILED )
In the future.. we could add the tempest test name
and build_status:failed to determine how often
certain tempest tests are failing.
This repository should not make use of openstack containt files and addresses sanity job failure
caused by switch of nodeset image used for testing.
We encountered an issue when repoquery took several minutes
which has lead to timeouts and unfinished log collections.
To avoid that this commit adds a timeout which is basically
a fail-safe from collect_logs point of view.
The patch replaces for loop for record_available_packages command
by a single command.
There have been problems with centos7 lately with py2/py3.
Ansible's pip module used python2 instead of python3 for
creating virtual venv which lead to dependency issues with
infrared which should be installed only by py3.
This changes also bumps the linters, removing yamllint which is now
included in ansible-lint.
- converts lonely unit test to use the official unittest format
for ansible collections.
- adds two tests to sova module
- moves sovalib into sova module as this is required in order to
make the module compatible with both role and collection deployments.
- upgrade linters
- enable black formatting so we don't waste time making flake8 happy
- all the .py files modified by this patch were modified by black itself
Because Ansible official testing tools (ansible-test) cannot be used
without a collection, we change the code layout to make conformant.
WARNING: The role is no longer considered to be named
"ansible-role-collect-logs" but "collect_logs" instead, with a
temporary alias called "collect-logs".
- [x] ansible-test sanity checks runs (does not need to pass)
- [x] zuul is still able to use the role
- [x] infrared is still able to use the role
- [x] molecule tests are running and passing
- [x] tripleo-ci jobs still collect the files
One symlink is still needed for infrared until related patch lands:
If the stack's name is not 'overcloud' then its rc file is not
collected. Rather than relying on a hard-coded stackrc,
overcloudrc, etc we can use a wildcard to catch them all. This
might also catch a few other files, but that's a small price to
None of our molecule tests were running but the job was executed.
This includes a fail-safe guard that should prevent such accidents
in the future. This assures the expected number of tests passed.
If /usr/bin/python is not available when trying to collect
the logs, then none are collected. In this patch we ensure
that alternative python interpreters are considered when
attempting the log collection. The interpreters are
considered in order of preference, with the system python
being the last and the default when the host is unreachable.
This resolves the issue when infrared has an interpreter
set in the inventory, but it's the wrong one due to the
host recently being upgraded from RHEL7 to RHEL8.
Since we use the requirements from upper-constraints,
don't add upper cap to 1.3 which is too old version and
u-c is beyond that since pike.
For dnf module specific issues, where one module is getting disabled
and others getting enabled will give a better insight to debug
podman or package related issues.
Signed-off-by: Chandan Kumar (raukadah) <email@example.com>