Change the way to estimate running components

[todor-ivanov]

Fixes #12184

Status

ready

Description

With the current PR we address the issue reported by T0 team in the following comment: #12184 (comment)

Here we change the way we estimate the set of running components. Instead of just listing the log directories here we rely on the returned value from the actual manage command and the execute-agent call to the underlying daemon to run the agent.

Is it backward compatible (if not, which system it affects?)

YES

Related PRs

#12185

External dependencies / deployment changes

None

[dmwm-bot]

Jenkins results:

• Python3 Unit tests: failed
  • 1 new failures
  • 2 changes in unstable tests
• Python3 Pylint check: succeeded
• Pycodestyle check: succeeded

Details at https://cmssdt.cern.ch/dmwm-jenkins/view/All/job/WMCore-PR-Report/110/artifact/artifacts/PullRequestReport.html

[todor-ivanov]

Thanks @vkuznet

[amaltaro]

With this change, we no longer look at components that have been terminated and no longer running.
Relying on the Running stdout string isn't reliable, as that is actually a bug that we have to eventually fix in the system: #12091

IMO, if the agent is supposed to be running, this script should be ensuring that any component down or stuck gets restarted - which was the previous behavior, unless I missed something subtle in these changes.