Skip to content
Snippets Groups Projects
Merged Sergey Karatkevich requested to merge kevit/ansible:master into master
1 unresolved thread

MR based on discussion

Merge request reports

Loading
Loading

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
34 34 name: "{{ item }}"
35 35 enabled: true
36 36 daemon_reload: true
37 daemon_reexec: yes
  • As far as I understand, this reexecute the whole systemd init process, isn't it a bit heavy handed?

  • Author Contributor

    Yes, it is, but otherwise systemd changes not getting picked up and services flapping after initial ansible run. I can share my test environment if you want to check by yourself how it looks like

  • Instead, maybe we could add a

    notify:
        - restart funkwhale

    handler to ensure the service is restarted on configuration change? (At least it wouldn't affect the global init daemon)

  • Author Contributor

    That is exactly the issue, while daemon_reload: true rereading systemd unit configurations and task herself start units properly. But inside units you can find OnFault directives, which push units to restart automatically when they in fault status. Unfortunately for some internal systemd reason you need to reserialize systemd configuration to make it work properly, probably because particular systemd units have dependencies to others. That is my understanding of situation at this point of time.

  • Author Contributor

    by the way, if you restart target or units manually it does not help as well

  • Thank you for the clarification. I've never encountered this bug myself and it never has been never reported before, so before merging a change that can potentially cause downtime on unrelated services, I'd really like to investigate if other persons have encountered this and if we can find another way.

    There's no reason to not merge your jinja fix though, so do you agree if we remove this second commit, merge the current PR to fix the HTTP2 issue, and open a separate issue to discuss the systemd issue?

  • Author Contributor

    sure

  • changed this line in version 3 of the diff

  • Thanks, I've merged your jinja change, we can can continue the discussion at opened #2 and, I'm especially interested in a way to reproduce the issue :)

  • Please register or sign in to reply
  • Thank you for the fix, ready to merge as soon as the systemd related part is clarified :)

  • Agate added 1 commit

    added 1 commit

    Compare with previous version

  • Agate mentioned in commit b08e97b9

    mentioned in commit b08e97b9

  • merged

  • Agate mentioned in issue #2

    mentioned in issue #2

  • Please register or sign in to reply
    Loading