[jira] [Commented] (HBASE-30014) refresh_meta not working due to regionNode lock

["Shanmukha Haripriya Kota (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-30014?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18068388#comment-18068388 ] 

Shanmukha Haripriya Kota commented on HBASE-30014:
--------------------------------------------------

HBASE-28158 moved RIT maintenance out of TRSP into RegionInTransitionTracker. Regions are added/removed from that tracker when handleRegionStateNodeOperation runs for the current RegionStateNode state. 
createAssignProcedure (called by RefreshMetaProcedure) attaches an assign TRSP with setProcedure(...) but, before the fix, did not invoke handleRegionStateNodeOperation at that moment. So there can a window of inconsistent behavior when listing the tables after refresh meta.

> refresh_meta not working due to regionNode lock
> -----------------------------------------------
>
>                 Key: HBASE-30014
>                 URL: https://issues.apache.org/jira/browse/HBASE-30014
>             Project: HBase
>          Issue Type: Sub-task
>          Components: read replicas, Replication
>            Reporter: Kevin Geiszler
>            Assignee: Shanmukha Haripriya Kota
>            Priority: Major
>              Labels: pull-request-available
>         Attachments: unable-to-acquire-lock-for-regionMode-replica-master.log, working-refresh-meta-master-replica.log
>
>
> There is an issue in our new "HBASE-29081_rebased" branch after rebasing with the master branch. I am seeing the following being repeated in master log after running {{refresh_meta}} on the replica cluster:
> {code:java}
> 2026-03-20T22:13:37,693 WARN  [RpcServer.priority.RWQ.Fifo.write.handler=2,queue=0,port=16000] assignment.AssignmentManager: Unable to acquire lock for regionNode state=OPEN, location=hbase-docker-2,16020,1774044759623, table=test_table, region=4f345a940e95b6eba65d80f1ec34d71c. It is likely that another thread is currently holding the lock. To avoid deadlock, skip execution for now.{code}
> I see this on the read-replica docker container setup. The steps to reproduce are:
>  # Start the hbase-docker container as the active cluster and hbase-docker-2 as the replica cluster
>  # On the active cluster, use the HBase shell to create a table
>  # On the replica cluster, run {{{}refresh_meta{}}}.  You will then see the "Unable to acquire lock for regionNode..." log message repeated in {{{}/opt/hbase/logs/hbase-jboss-master-hbase-docker-2.log{}}}. Also, after running {{{}refresh_meta{}}}, you will not see the table that was created on the active cluster.
> I have attached a log from both the working branch (HBASE-29081) and the branch where it does not work (HBASE-29081_rebased).  Search the logs for "Submitting RefreshMetaProcedure" to see where the {{refresh_meta}} procedure began.



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