Please take a jstack next time. What version of jvm are you using. If older, < u21, and no apparent reason for the RegionServer lockup, then it might be a JVM issue. Try adding -XX:+UseMembar to your HBASE_OPTS (We've seen at least one case of RegionServer lockup because of missed notifications).
St.Ack On Wed, Mar 23, 2011 at 4:23 AM, Iulia Zidaru <[email protected]> wrote: > Hi all, > > We are using CDH3B4. > The RegionServer containing .META. stopped working under normal load. The > region server process was up, it could be viewed in hbase admin interface, > but the requests, regions, usedHeap, maxHeap metrics were constant in time > and it couldn't be accessed. Other machines continued working, but the > performance was very bad. A huge log was created on the machine's HDFS. The > cluster came back to normal performance when I had killed the RegionServer. > In regionServer's log we got only a WARNING > (java.nio.channels.ClosedChannelException) cause by the communication with > the client machine. No logs while it was frozen. > Any idea? > > Thank you, > Iulia > > > -- > Iulia Zidaru > Java Developer > > 1&1 Internet AG - Bucharest/Romania - Web Components Romania > 18 Mircea Eliade St > Sect 1, Bucharest > RO Bucharest, 012015 > [email protected] > 0040 31 223 9153 > > >
