oc, sounds correct, but why would you need to reduce the number of retries? usually when there is a transient failure in the ZK quorum (e.g. network partition) you would rely on this setting for the HBase Region Servers to retry connections to ZK until ZK is one more time available or it runs of retries. Probably it might be better idea to add more ZK severs (3 in total ideally) than lower the number of retries or find the root cause why the RS or the NN went down.
cheers, esteban. -- Cloudera, Inc. On Tue, Jul 29, 2014 at 8:53 AM, oc tsdb <[email protected]> wrote: > Hi, > > Thanks for your response. > > We are using Hbase 0.98.2. > > You mean we need to add the following two properties in hbase-site.xml? > zookeeper.recovery.retry = 1 > zookeeper.recovery.retry.intervalmill=1000 > > Thanks > oc.tsdb > > > > On Tue, Jul 29, 2014 at 8:09 PM, Ted Yu <[email protected]> wrote: > > > You can adjust values for the following config parameters (values below > are > > only for illustration): > > > > config.setInt("zookeeper.recovery.retry", 1); > > config.setInt("zookeeper.recovery.retry.intervalmill", 1000); > > > > What hbase release are you using ? > > > > Cheers > > > > > > On Tue, Jul 29, 2014 at 3:48 AM, oc tsdb <[email protected]> wrote: > > > > > Hi, > > > > > > if the namenode (or zookeeper) is down for some reason, > > > hbase-regionserver(on datanode) keep retrying to connect to zookeeper > > > continuously for long time. > > > > > > We tried by setting "hbase.client.retries.number" to 1 in > hbase-site.xml, > > > But no effect. > > > > > > Can someone please let us know, hbase configuration to stop > > > hbase-regionserver in 5 sec, in case, if zookeeper is not running. > > > > > > Thanks > > > oc.tsdb > > > > > >
