Actions
Backport #56152
closedpacific: mgr/snap_schedule: schedule updates are not persisted across mgr restart
Release:
pacific
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
https://github.com/ceph/ceph/pull/46797
scrub status does not reflect the correct status after mgr restart
eg.
Before restart mgrs:
[{"fs": "cephfs", "subvol": null, "path": "/", "rel_path": "/", "schedule": "1h", "retention": {"h": 6}, "start": "2022-06-07T12:50:00", "created": "2022-06-07T16:44:25", "first": "2022-06-07T16:50:00", "last": "2022-06-07T16:50:00", "last_pruned": null, "created_count": 1, "pruned_count": 0, "active": true}] $ ceph fs snap-schedule status / --format=json
After Restart mgrs:
[{"fs": "cephfs", "subvol": null, "path": "/", "rel_path": "/", "schedule": "1h", "retention": {}, "start": "2022-06-07T12:50:00", "created": "2022-06-07T16:44:25", "first": null, "last": null, "last_pruned": null, "created_count": 0, "pruned_count": 0, "active": true}] $ ceph fs snap-schedule status / --format=json
Need for this backport only tracker: ------------------------------------ * This issue needs to be fixed exclusively in the pacific branch since the db operations take place on the in-memory sqlite db and are not automatically persisted to permanent storage. * This db management policy is older and different than the current mainline. In the mainline sources, the sqlite db uses the ceph backend to persist db changes directly to RADOS objects. * Although there is a mechanism to write the in-memory db to the stable storage, the calls to the appropriate procedure needs to be added to functions which update the state of the db.
Actions