Ceph RADOS Gateway Charm

changing source doesn't restart <email address hidden>

Bug #1934935 reported by Drew Freiberger
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Ceph RADOS Gateway Charm
New
Undecided
Unassigned

Bug Description

On openstack charms version 21.04, when upgrading the ceph-radosgw service with "juju config ceph-radosgw source=cloud:bionic-rocky", the rgw service stays running under luminous instead of restarting the service on mimic.

This can be reproduced by deploying ceph-radosgw on bionic with source=distro, then running above juju config command, and finally, running "juju run -u ceph-mon/leader 'ceph versions'" and noting that the "rgw" service version is still showing 12.2.13 luminous (stable)".

A simple workaround after the upgraded packages are installed is to run:

juju run -a ceph-radosgw 'systemctl restart ceph-radosgw.target'

witnessed:
    "rgw": {
        "ceph version 12.2.13 (584a20eb0237c657dc0567da126be145106aa47e) luminous (stable)": 3
    },

expected:
    "rgw": {
        "ceph version 13.2.8 (5579a94fafbc1f9cc913a0f5d362953a5d9c3ae0) mimic (stable)": 3
    },

Drew Freiberger (afreiberger)
tags: added: openstack-upgrade
Revision history for this message
Drew Freiberger (afreiberger) wrote :

Maybe related to ceph-mgr bug of similar nature: https://bugs.launchpad.net/charm-ceph-mon/+bug/1849874

Revision history for this message
Drew Freiberger (afreiberger) wrote :

This seems to be an issue during any package upgrade of radosgw.

The radosgw service is masked, and the actual service to restart is ceph-radosgw@rgw.<hostname>.service.

It would be helpful if the deb packaging post-install script could determine if this @rgw.<hostname> service is registered and restart it instead of radosgw.

Revision history for this message
Alex Kavanagh (ajkavanagh) wrote :

Hi Drew, I think this is a dup of https://bugs.launchpad.net/charm-ceph-radosgw/+bug/1906707 - if it is, it may be better to combine all the info and mark this as a dup? I think this bug report has more useful info, though.

Revision history for this message
Drew Freiberger (afreiberger) wrote :

I agree that this bug is a duplicate of the other, however, in comment #2 it's not actually an openstack-upgrade bug, but a package upgrade bug. the package doesn't restart the service after upgrade, rather than needing to fix the charm, I think we need to address the packaging issue such that when patches are made available in the repository of a single openstack version that the @rgw.hostname service is restarted after package (rather than just openstack) upgrade.

Revision history for this message
Billy Olsen (billy-olsen) wrote :

The ceph packages do not automatically restart the ceph services, by design. This is to avoid unplanned disruption across the storage platform. This includes osds, mons, rgws, mds, etc. The charms should handle the restarts appropriately. As such, I'm marking this as a duplicate.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.