hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From mail list <louis.hust...@gmail.com>
Subject Slow waitForAckedSeqno took too long time
Date Fri, 05 Dec 2014 05:01:23 GMT
Hi ,all

I deploy Hbase0.98.6-cdh5.2.0 on 3 machine:

l-hbase1.dev.dba.cn0(hadoop namenode active, HMaster active)
l-hbase2.dev.dba.cn0(hadoop namenode standby, HMaster standby, hadoop datanode)
l-hbase3.dev.dba.cn0(regionserver, hadoop datanode)

Then I shutdown the l-hbase1.dev.dba.cn0,  But HBase can not work until about 15mins later.
I check the log and find the following log in the region server’s log:

2014-12-05 12:03:19,169 WARN  [regionserver60020-WAL.AsyncSyncer0] hdfs.DFSClient: Slow waitForAckedSeqno
took 927762ms (threshold=30000ms)
2014-12-05 12:03:19,186 INFO  [regionserver60020-WAL.AsyncSyncer0] wal.FSHLog: Slow sync cost:
927779 ms, current pipeline: []
2014-12-05 12:03:19,186 DEBUG [regionserver60020.logRoller] regionserver.LogRoller: HLog roll
2014-12-05 12:03:19,236 WARN  [regionserver60020-WAL.AsyncSyncer1] hdfs.DFSClient: Slow waitForAckedSeqno
took 867706ms (threshold=30000ms)

It seems the WAL Asysnc took too long time for region server recovery? I don’t know if the
log matters ?
Can any body explain the reason? and how to reduce the time for recovery?

View raw message