Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
Elixir, 7.1.2
-
Untriaged
-
1
-
Unknown
Description
When the cluster is re-run, the cbauth database is stale and hence the ctl.StartCtl call in main.go returns an error which is logged as fatal. This leads to a PublishCrashEvent which crashes the cbft process because of a nil pointer access which leads to a restart of the cbft process. This cycle occurs multiple times until the cbauth is fresh and doesn't reject the requests.
Attachments
| For Gerrit Dashboard: MB-52853 | ||||||
|---|---|---|---|---|---|---|
| # | Subject | Branch | Project | Status | CR | V |
| 177132,2 | MB-52853 - cbft process crash and restart multiple times | neo | cbgt | Status: MERGED | +2 | +1 |
| 177213,1 | Merge remote-tracking branch 'couchbase/neo' | master | cbgt | Status: MERGED | +2 | +1 |
Activity
Build couchbase-server-7.1.2-3356 contains cbgt commit 4d792f1 with commit message:
MB-52853 - cbft process crash and restart multiple times
Build couchbase-server-7.2.0-1563 contains cbgt commit 4d792f1 with commit message:
MB-52853 - cbft process crash and restart multiple times
Build couchbase-server-8.0.0-1047 contains cbgt commit 4d792f1 with commit message:
MB-52853 - cbft process crash and restart multiple times
Thejas Orkombu Does this needs QE verification. if yes, can you please update with reproducible steps?
In my opinion, its not needed. But the reproducible steps are (on a build without this fix)
1. Run a fresh cluster, avail fts service, do fts related operations if needed.
2. stop the cluster and run it again.
3. In the logs, you'd observe crash messages due to a nil pointer access.
With this fix, the crash is avoided in the system event code path.
As per below comment. Closing the ticket
Thejas Orkombu added a comment - 2 days ago
Just an FYI, this issue has been dev-verified.
Wayne Siu an unhandled error causes a seg fault and crashes the cbft, due to stale cbauth. Please approve the ticket for 7.1.2.