You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Duo Zhang (Jira)" <ji...@apache.org> on 2024/07/18 03:55:00 UTC

[jira] [Commented] (HBASE-28582) ModifyTableProcedure should not reset TRSP on region node when closing unused region replicas

    [ https://issues.apache.org/jira/browse/HBASE-28582?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17866909#comment-17866909 ] 

Duo Zhang commented on HBASE-28582:
-----------------------------------

Ah, when implementing HBASE-28522, I found that there is a very big problem that, in CloseExcessRegionReplicasProcedure, I forgot to call super.serializeStateData...

This is a very critical problem as it may cause the loading of procedure store to fail and bring down the whole cluster...

[~apurtell] I think we need to make a 2.5.9.1 release to include an addendum for this fix...

> ModifyTableProcedure should not reset TRSP on region node when closing unused region replicas
> ---------------------------------------------------------------------------------------------
>
>                 Key: HBASE-28582
>                 URL: https://issues.apache.org/jira/browse/HBASE-28582
>             Project: HBase
>          Issue Type: Bug
>          Components: proc-v2
>            Reporter: Duo Zhang
>            Assignee: Duo Zhang
>            Priority: Critical
>              Labels: pull-request-available
>             Fix For: 2.7.0, 3.0.0-beta-2, 2.6.1, 2.5.9
>
>
> Found this when digging HBASE-28522.
> First, this is not safe as MTP does not like DTP where we hold the exclusive lock all the time.
> Second, even if we hold the exclusive lock all the time, as showed in HBASE-28522, we may still hang there forever because SCP will not interrupt the TRSP.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)