"Protocol violation. JOIN message sender ... is not in state transfer (SYNCED). Message ignored." with concurrent IST

[Philip Stoev]

A test that performs concurrent IST with two joiners and one donor, produces semi-reliably with the following warning:

2014-10-10 16:37:31 19225 [Warning] WSREP: Protocol violation. JOIN message sender 0.0 (localhost.localdomain) is not in state transfer (SYNCED). Message ignored.

[Philip Stoev (philip-stoev-f)]

Attaching a test case. Please unpack in mysql-test and run:

 perl mysql-test-run.pl --suite galera_3nodes galera_simultaneous_ist --big-test

[Teemu Ollakka (teemu-ollakka)]

Corresponding Github issue: https://github.com/codership/galera/issues/101

[Alex Yurchenko (ayurchen)]

or, rather, https://github.com/codership/galera/issues/106, cause all work is done there

Philip, does it happen with 106 branch?

[Philip Stoev (philip-stoev-f)]

Yes, I believe the gh106 branch is also affected . I will retest on Monday. Also, Yan told me that the master and the gh106 branch are currently identical.

[Philip Stoev (philip-stoev-f)]

For some reason, I can no longer reproduce it on gh106. It is still reproducible on master.

If you feel that this bug is therefore a dead end, please close it when gh106 is pushed.

[Raghavendra D Prabhu (raghavendra-prabhu)]

@Phillip,

Is this reproduceable with master now?

[Philip Stoev (philip-stoev-f)]

The warning is now:

2014-12-15 10:53:26 63468 [Warning] WSREP: SYNC message from member 0 in non-primary configuration. Ignored.

[Philip Stoev (philip-stoev-f)]

Teemu reports that the warning currently produced by the test requires further investigation.

[Philip Stoev (philip-stoev-f)]

The new warning is now

https://github.com/codership/galera/issues/204

I am closing this ticket because the previous warning is no longer reproducible.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PXC-1752