Solve KCM failure on system tests on F40 #69
Conversation
|
Thank you. We disable burst limit for sssd at https://github.com/SSSD/sssd-test-framework/blob/master/sssd_test_framework/hosts/client.py#L74 so the same can be done for KCM, but this looks like a better approach. Can you remove the burst limit configuration? I think we can move reset_failed directly to pytest-mh, e.g. call It should only be needed for start and restart, not stop. |
Done.
Done too. |
|
Thanks. I added commit to bump pytest-mh minimum version. |
When SSSD was synchronously started, the KCM service was stopped. When used, KCM would be started by the socket unit, allowing it to re-read the configuration. This behavior was not coherent across start(), stop(), async_start(), async_stop(), restart() and async_restart(). This commits makes the behavior coherent across these methods.
Now that pytest-mh:SystemdServices is reseting the failed flag, this is no longer needed.
Sudo tests are failing as a result of KCM not being started.
systemdrefuses to start KCM because some tests are starting and stopping KCM too quickly andsystemdconsiders this is a failure.To avoid this behavior,
pytest-mhnow resets the failed flag every time a unit is started. Because this behavior was changed specifically for this fix, we need to get next-actions/pytest-mh#34 pulled before pulling this one.in addition, we made the handling of KCM coherent through the
start(),stop(),restart()and their asynchronous versions. So far onlystart()would stop KCM for it to re-read its configuration.Because of the changes in
pytest-mh, disabling the bust limit is no longer necessary and was removed.