[
https://issues.apache.org/jira/browse/HBASE-20817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16527459#comment-16527459
]
Duo Zhang commented on HBASE-20817:
-----------------------------------
{noformat}
2018-06-29 16:56:38,591 INFO
[RpcServer.priority.FPBQ.Fifo.handler=5,queue=0,port=37605]
regionserver.RSRpcServices(1607): Close 31399d62546fa468268ed297bfdb37fa,
moving to zhangduo-ubuntu,44909,1530262591610
2018-06-29 16:56:38,594 DEBUG
[RS_CLOSE_REGION-regionserver/zhangduo-ubuntu:0-0] regionserver.HRegion(1527):
Closing 31399d62546fa468268ed297bfdb37fa, disabling compactions & flushes
2018-06-29 16:56:38,595 DEBUG
[RS_CLOSE_REGION-regionserver/zhangduo-ubuntu:0-0] regionserver.HRegion(1567):
Updates disabled for region
test,,1530262597448.31399d62546fa468268ed297bfdb37fa.
2018-06-29 16:56:38,605 DEBUG
[RS_CLOSE_REGION-regionserver/zhangduo-ubuntu:0-0] wal.WALSplitter(678): Wrote
file=hdfs://localhost:45953/user/zhangduo/test-data/e118edb1-2859-af77-4b1b-447b432b1f10/data/default/test/31399d62546fa468268ed297bfdb37fa/recovered.edits/4.seqid,
newMaxSeqId=4, maxSeqId=1
2018-06-29 16:56:38,607 INFO
[RS_CLOSE_REGION-regionserver/zhangduo-ubuntu:0-0] regionserver.HRegion(1681):
Closed test,,1530262597448.31399d62546fa468268ed297bfdb37fa.
2018-06-29 16:56:38,607 INFO
[RS_CLOSE_REGION-regionserver/zhangduo-ubuntu:0-0]
regionserver.HRegionServer(3426): Adding 31399d62546fa468268ed297bfdb37fa move
to zhangduo-ubuntu,44909,1530262591610 record at close sequenceid=2
{noformat}
[~elserj] Write a UT to move region, this is the log when closing the region.
You can see that there is a log written by WALSplitter about writing max
sequence id file. Could you please enable the DEBUG log to see if there is one
in your cluster?
{code}
// Store SeqId in HDFS when a region closes
// checking region folder exists is due to many tests which delete the
table folder while a
// table is still online
if (this.fs.getFileSystem().exists(this.fs.getRegionDir())) {
WALSplitter.writeRegionSequenceIdFile(this.fs.getFileSystem(),
this.fs.getRegionDir(),
mvcc.getReadPoint());
}
{code}
As you can see that, the check is a dummy one, as on real cluster, the region
directory should always be presented...
> Infinite loop when executing ReopenTableRegionsProcedure
> ---------------------------------------------------------
>
> Key: HBASE-20817
> URL: https://issues.apache.org/jira/browse/HBASE-20817
> Project: HBase
> Issue Type: Bug
> Components: Region Assignment
> Reporter: Duo Zhang
> Priority: Blocker
> Fix For: 3.0.0, 2.1.0, 2.0.2, 2.2.0
>
>
> As discussed in HBASE-20792, it seems that a region's openSeqNum could remain
> the same after a sucessful reopen, which causes the RTRP loop infinitely.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
