Bug #41823
[rbd-mirror] service daemon instance id might not match
Status:
New
Priority:
Normal
Assignee:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
luminous,mimic,nautilus
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
It's possible for the MONs to reset the connection's global unique id. In that case, the output from "ceph service status" shows the inconsistency:
# ceph service status { "rbd-mirror": { "84243": { "status_stamp": "2019-09-13 15:40:01.149815", "last_beacon": "2019-09-13 15:40:26.151381", "status": { "json": "{\"2\":{\"name\":\"rbd\",\"callouts\":{},\"image_assigned_count\":51,\"image_error_count\":0,\"image_local_count\":51,\"image_remote_count\":51,\"image_warning_count\":0,\"instance_id\":\"84247\",\"leader\":true}}" } } } }
(i.e. 84243 vs 84247)
History
#1 Updated by Jason Dillaman over 4 years ago
Correction: the service daemon instance id is from a different cluster connection than the per-pool connections established for replaying. There is still a potential for an issue should the MON reset the instance id, but it appears to be a much less worrisome issue.