osd/scrub: do not start scrubbing if the PG is snap-trimming #46440
Conversation
|
--- Will modify the commit message to include the reason for not allowing snaps being removed underneath a running scrub. --- Done. |
ronen-fr
force-pushed
the
wip-rf-scrub-trimisbad
branch
from
b157eb7
to
0a6156f
Compare
|
jenkins test api |
There was a problem hiding this comment.
Is this issue already tracked in https://tracker.ceph.com/issues/49525? Otherwise, LGTM
|
"Is this issue already tracked in https://tracker.ceph.com/issues/49525? Otherwise, LGTM" |
|
@neha-ojha - candidate for Quincy backport? |
Both 'snap-trim' and 'snaptrim-wait' PG states now prevent scrub from starting. Background: A PG should not be scrubbed and trimmed concurrently. Unlike write operations, snap trimming does not verify that a targeted object is not in the executing scrub's chunk. The trimmer always checked for active scrubs before starting. The scrubber - did not. This PR fixes that omission. Fixes: https://tracker.ceph.com/issues/49525 Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
ronen-fr
force-pushed
the
wip-rf-scrub-trimisbad
branch
from
0a6156f
to
51816cd
Compare
yes, https://tracker.ceph.com/issues/49525 was reported a more than a year ago, worth checking if pacific also needs the fix |
The problem surely exists in Pacific. Actually - as far as I checked, the same omission exists in the original scrub code. |
|
Merging based on http://pulpito.front.sepia.ceph.com/yuriw-2022-06-01_23:19:00-rados-wip-yuri8-testing-2022-06-01-1114-distro-default-smithi. |
|
Looks the quincy backport is: #46498. |
Both 'snap-trim' and 'snaptrim-wait' PG states will now prevent
scrub from starting.
Fixes: https://tracker.ceph.com/issues/49525
Signed-off-by: Ronen Friedman rfriedma@redhat.com