Hi, You are definitely experiencing performance degradation cussed by major compaction. How big is your write load ? Maybe you should consider to decrease time interval between major compactions if you are writing a lot data to cluster. Also here is part of hbase book which explains how to configure major compaction related parameters:
http://hbase.apache.org/book/regions.arch.html#compaction Regards Samir On Tue, Jun 17, 2014 at 9:16 AM, yanivG <[email protected]> wrote: > We are running major compaction once a week, on a 6 nodes hbase cluster. > We see that we are getting timeouts in our application. > We see also in JMX metrics the following: > 1. compaction time avg was 200 seconds instead of 3 seconds > 2. ParNew collection time was twice than usual > 3. Compaction queue size was high > 4. Region number increased at that time > 5. Disk I\O read\write was high > > Can we overcome it by tuning hbase to be more patient during major > compaction? > > > > -- > View this message in context: > http://apache-hbase.679495.n3.nabble.com/Timeouts-during-major-compaction-tp4060507.html > Sent from the HBase User mailing list archive at Nabble.com. >
