[
https://issues.apache.org/jira/browse/HBASE-29359?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17955505#comment-17955505
]
Duo Zhang commented on HBASE-29359:
-----------------------------------
I think this is intentional.
Please see the commit for HBASE-26874, they will wait in cleanup method if
there are unfinished recompare tasks. They do not want to wait in the main map
loop method as it will greatly slow down the compare job.
> VerifyReplication needs to join on its verification tasks
> ---------------------------------------------------------
>
> Key: HBASE-29359
> URL: https://issues.apache.org/jira/browse/HBASE-29359
> Project: HBase
> Issue Type: Bug
> Components: Replication
> Reporter: Nick Dimiduk
> Assignee: Nick Dimiduk
> Priority: Major
> Labels: pull-request-available
>
> Using the VerifyReplication job, I observed many reported inconsistencies
> between the tables. Inspecting the logs, I see lots of
> {noformat}
> WARN o.a.h.h.m.r.VerifyReplicationRecompareRunnable - Sleeping interrupted,
> incrementing bad rows and aborting
> {noformat}
> I think the main map task needs to join on the worker threads it spawns.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
