[
https://issues.apache.org/jira/browse/HBASE-2018?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12785486#action_12785486
]
stack commented on HBASE-2018:
------------------------------
+1 on v2 of patch.
> Updates to .META. blocked under high MemStore load
> --------------------------------------------------
>
> Key: HBASE-2018
> URL: https://issues.apache.org/jira/browse/HBASE-2018
> Project: Hadoop HBase
> Issue Type: Bug
> Affects Versions: 0.20.2
> Reporter: Jean-Daniel Cryans
> Assignee: Jean-Daniel Cryans
> Priority: Blocker
> Fix For: 0.20.3, 0.21.0
>
> Attachments: HBASE-2018-v2.patch, HBASE-2018.patch
>
>
> I discovered this on Lars' cluster. The symptom was the good old:
> {code}
> 09/11/30 08:10:26 INFO mapred.JobClient: Task Id :
> attempt_200911250121_0011_r_000010_1, Status : FAILED
> org.apache.hadoop.hbase.client.RetriesExhaustedException: Trying to contact
> region server Some server, retryOnlyOne=true, index=0, islastrow=false,
> tries=9, numtries=10, i=14, listsize=20,
> region=prev-docs,de68fb97795ef3d936a3f10ff8790253,1259573366564 for region
> prev-docs,ccea967e66ccb53d83c48849c3a23f21,1259542138868, row
> 'ccff8cd4ca871c41f4fa7d44cffed962', but failed after 10 attempts.
> Exceptions:
> at
> org.apache.hadoop.hbase.client.HConnectionManager$TableServers$Batch.process(HConnectionManager.java:1120)
> at
> org.apache.hadoop.hbase.client.HConnectionManager$TableServers.processBatchOfRows(HConnectionManager.java:1201)
> at org.apache.hadoop.hbase.client.HTable.flushCommits(HTable.java:605)
> at org.apache.hadoop.hbase.client.HTable.put(HTable.java:470)
> at org.apache.hadoop.hbase.mapreduce.TableOutputFormat$TableRecordW
> {code}
> But the load wasn't that heavy, just lots of splitting going on. Looking at
> the logs, I see a split taking more than 4 minutes which is explained by this
> happening on the RS hosting .META. :
> {code}
> 2009-11-30 08:08:39,922 INFO
> org.apache.hadoop.hbase.regionserver.MemStoreFlusher: Forced flushing of
> prev-docs,2c9d51e57b20decd5c6419d23ede822b,1259542273901 because global
> memstore limit of 1.6g exceeded; currently 1.6g and flushing till 1021.9m
> ...
> 2009-11-30 08:12:33,743 DEBUG org.apache.hadoop.hbase.regionserver.HRegion:
> Finished memstore flush of ~22.9m for region
> prev-docs,c8fea4fbbc41e746d960854ed4d41dd6,1259587143838 in 14160ms, sequence
> id=13677, compaction requested=false
> 2009-11-30 08:12:33,744 INFO
> org.apache.hadoop.hbase.regionserver.MemStoreFlusher: Forced flushing of
> prev-docs,39c2995d955c041d21f4dc4a0d0dbf6c,1259587061295 because global
> memstore limit of 1.6g exceeded; currently 1.0g and flushing till 1021.9m
> {code}
> So we should not block updates to .META. for any reason. I'm pretty sure this
> issue explains other issues we've seen on the mailing list.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
