[MXS-3712] Maxscale not adjusting event status after failover in skysql Created: 2021-08-04  Updated: 2021-10-11  Resolved: 2021-10-08

Status: Closed
Project: MariaDB MaxScale
Component/s: N/A
Affects Version/s: 2.5.3
Fix Version/s: 2.5.14

Type: Bug Priority: Blocker
Reporter: Isaac Venn (Inactive) Assignee: Unassigned
Resolution: Fixed Votes: 0
Labels: triage


 Description   

When Maxscale is automatically failing over in skysql, the new master is not having the events updated to "enabled" from "slaveside_disabled". As a result, clients are reporting to us that events are breaking. There is concern that not only is the even not firing on the master anymore, it may be firing on the replica making the data out of sync.
Reported originally from whitebox where they are running mariadb 10.5.5-3



 Comments   
Comment by markus makela [ 2021-08-04 ]

Could this be related to MXS-3158?

Comment by markus makela [ 2021-08-25 ]

ivenn does this affect newer versions of MaxScale?

Comment by markus makela [ 2021-09-14 ]

I'll close this as Incomplete since it's a very old version and we haven't received any feedback. If you see this again, let us know and we'll reopen this.

Comment by markus makela [ 2021-09-29 ]

Reopening as the original problem still remains.

Comment by Isaac Venn (Inactive) [ 2021-10-06 ]

Unfortunately this is still occurring. What is happening is when maxscale fails over the master to a replica, events that were set as "slaveside_disabled" are not being updated to "enabled" and don't fire. What's worse is that the previous master is not being updated to "slaveside_disabled" so the data is getting out of sync as it only fires on a single replica and not the master.

Comment by markus makela [ 2021-10-06 ]

Can you update the MaxScale version you saw this with?

Comment by markus makela [ 2021-10-06 ]

Additionally, if it's possible to capture the log output from MaxScale when this happens, it would greatly help us resolve this.

Comment by Isaac Venn (Inactive) [ 2021-10-06 ]

The version was 2.5.3. I do not have logs on it at this time. I'd have to spin up a skysql cluster and force the failure.

Comment by markus makela [ 2021-10-07 ]

Have you ever observed this with a newer MaxScale version? Is it possible for you to test with something newer than 2.5.3?

Comment by Isaac Venn (Inactive) [ 2021-10-07 ]

This does appear to be working in 2.5.14 (which is currently in skysql) so I believe this can be closed.

Generated at Thu Feb 08 04:23:24 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.