Actions
Bug #26966
closednfs-ganesha: epochs out of sync in cluster
% Done:
0%
Source:
Development
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
Client, Ganesha FSAL
Labels (FS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
After starting a brand new nfs-ganesha cluster, I found the following set of recovery dbs:
$ rados -p nfs-ganesha ls rec-0000000000000002:a rec-0000000000000003:c rec-0000000000000003:b grace $ ganesha-rados-grace dump cur=3 rec=0 ====================================================== a b c
Two of them are at epoch 3, and the other is at epoch 2. I believe this probably happened because these hosts had no active clients, and so the last node needing a grace period cleared its flag and lifted the grace period before the first node (a) had noticed that a new grace period had started.
We may need to ensure that we can't lift the grace period until all nodes are enforcing.
Actions