Actions
Bug #57617
closedcrimson: need to actually set version/user_version for duplicate ops
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
Tends to create ceph_test_rados failures that look like:
... 2022-09-20T01:04:19.432 INFO:tasks.rados.rados.0.smithi182.stdout:74: oid 74 version 0 is already newer than 0 2022-09-20T01:04:19.432 INFO:tasks.rados.rados.0.smithi182.stdout:74: oid 74 updating version 0 to 19 2022-09-20T01:04:19.432 INFO:tasks.rados.rados.0.smithi182.stdout:74: oid 74 version 19 is already newer than 0 2022-09-20T01:04:19.432 INFO:tasks.rados.rados.0.smithi182.stdout:74: oid 74 version 19 is already newer than 0 2022-09-20T01:04:19.432 INFO:tasks.rados.rados.0.smithi182.stdout:update_object_version oid 74 v 19 (ObjNum 73 snap 0 seq_num 73) dirty exists 2022-09-20T01:04:19.433 INFO:tasks.rados.rados.0.smithi182.stderr:Error: racing read on 74 returned version 22 rather than version 19 ...
Note the "already newer than 0" bit -- the MOSDOpReply message we send when we detect a duplicate op has a field that should contain the user_version recorded in the pg log where we found the duplicate, but it's not being populated.
Actions