[jira] [Commented] (KAFKA-10719) MirrorMaker2 fails to update its runtime configuration

["Daniel Urban (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-10719?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17393104#comment-17393104 ] 

Daniel Urban commented on KAFKA-10719:
--------------------------------------

I believe that the core issue here is that the connectors will start up with the old config first, and then the new config gets applied by MM2.

If the old listener is not available, that will cause timeouts in the rebalance. Connectors have a default timeout which exceeds the default rebalance timeout. When the rebalance timeout happens, none of the MM2 nodes will be able to update the config after the startup.

Besides deleting the config topic, you can also try to increase the rebalance timeout of the Connect workers running inside MM2 (to something around 3 minutes), which will allow the cluster to not hit a rebalance timeout, and then successfully apply the new configs.

> MirrorMaker2 fails to update its runtime configuration
> ------------------------------------------------------
>
>                 Key: KAFKA-10719
>                 URL: https://issues.apache.org/jira/browse/KAFKA-10719
>             Project: Kafka
>          Issue Type: Bug
>    Affects Versions: 2.6.0
>            Reporter: Peter Sinoros-Szabo
>            Priority: Major
>
> I was running successfully the MM2 cluster with the following configuration, I simplified it a little: {code:java} clusters = main, backup main.bootstrap.servers = kafkaA:9202,kafkaB:9092,kafkaB:9092 backup.bootstrap.servers = backupA:9092,backupB:9092,backupC:9092 main->backup.enabled = true main->backup.topics = .*{code} I wanted to change the bootstrap.address list of the destination cluster to a different list that is pointing to the *same* cluster, just a different listener with a different routing. So I changed it to: {code:java} backup.bootstrap.servers = backupSecA:1234,backupSecB:1234,backupSecC:1234{code} I did a rolling restart on the MM2 nodes and say that some tasks were still using the old bootstrap addresses, some of them were using the new one. I don't have the logs, so unfortunately I don't know which one picked up the good values and which didn't. I even stopped the cluster completely, but it didn't help. Ryanne adviced to delete the mm2-config and mm2-status topics, so I did delete those on the destination cluster, that solved this issue.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)