osd/scrub: ignoring unsolicited DigestUpdate events#44050
osd/scrub: ignoring unsolicited DigestUpdate events#44050ronen-fr merged 1 commit intoceph:masterfrom
Conversation
|
Are we aware of a bug that causes this? I'm worried that this will hide problems with event ordering. Can we at least emit a central log warning and modify teuthology to fail any test that generates it? |
I agree to your suggestions to send the messages to the cluster log, and fail relevant tests. Will modify accordingly. |
155383d to
63e2873
Compare
|
I just want to make sure we actually notice if we start seeing these events again. |
|
This pull request can no longer be automatically merged: a rebase is needed and changes have to be manually resolved |
63e2873 to
fb2225e
Compare
fb2225e to
30a9486
Compare
30a9486 to
bf4d806
Compare
... and not just to the OSD's log Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
bf4d806 to
84fe362
Compare
|
I'll be merging this, based on Teuthology runs. |
|
jenkins test make check |
... and not just to the OSD's log Cherry-picking of PR ceph#44050. Manual editing was required to fix the WaitDigestUpdate state reactions, as the relevant change was missing in the picked commit (was pushed by mistake by a separate PR) (cherry picked from commit 84fe362) Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
DigestUpdate events arriving too early, i.e. in WaitReplicas state, can
and should be ignored. The down-counter for pending digest updates is
checked upon entering state WaitDigestUpdates, and a zero counter will be
noticed then. And if the early event is a result of a bug -
handling it is a mistake.
Signed-off-by: Ronen Friedman rfriedma@redhat.com