hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Yu <yuzhih...@gmail.com>
Subject Re: HBase 0.92/Hadoop 0.22 test results
Date Tue, 08 Nov 2011 17:33:06 GMT
> 11/11/08 00:43:40 INFO regionserver.HRegion: Finished memstore flush
> of ~18.4m for region
> TestLoadAndVerify_
1320729464658,<\xA1\xAF(k\xCA\x1A\xEA,1320729465485.8bd8387431feec2b09983693dfac950b.
> in 829ms, sequenceid=142942, compaction requested=false
> 11/11/08 00:44:31 INFO zookeeper.ClientCnxn: Unable to read additional
> data from server sessionid 0x133817270190001, likely server has closed
> socket, closing socket connection and attempting reconnect

Is there a way to find out what could have led to the ~1min gap above ?

Also, to help narrow our search, would HBase 0.92 + hadoop 0.20.205 produce
the
YouAreDeadException?

Thanks

On Tue, Nov 8, 2011 at 9:26 AM, Roman Shaposhnik <rvs@apache.org> wrote:

> On Tue, Nov 8, 2011 at 9:20 AM, Ted Yu <yuzhihong@gmail.com> wrote:
> > Roman:
> >> 11/11/08 00:44:31 WARN util.Sleeper: We slept 38891ms instead of
> >> 3000ms, this is likely due to a long garbage collecting pause and it's
> >> usually bad, see
> >
> > 3000ms is the default value for hbase.regionserver.msginterval
> > Obviously it is too short for the validation scenario.
> >
> > Can you increase its value and perform another round of test ?
>
> Sure, but I have always thought 3000 was long enough for a tiny
> cluster. We're not talking hundreds of nodes here. Has something
> in HBase architecture changed so that this value now needs to be
> bumped?
>  <property>
>    <name>hbase.regionserver.msginterval</name>
>    <value>1000</value>
>    <description>Interval between messages from the RegionServer to HMaster
>    in milliseconds.  Default is 15. Set this value low if you want unit
>    tests to be responsive.
>    </description>
>  </property>
> T
> hanks,
> Roman.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message