You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by hiroshi <hi...@isp21.co.jp> on 2009/12/29 05:03:24 UTC
zookeeper setting
Hi,
I built hadoop(0.20.1)+hbase(0.20.1) cluster structure(Master x 1, Slave x
4).
And I tried following process to verify the behavior during system
failure.Shut down one of the slave server, and start Hbase while I have a
machine which cannot establish connection in the cluster.
I this regard, I get the following log repeatedly(3-8times)at interval of 3
seconds on "hbase-hadoop-zookeeper-hmaster.log", as a result.
#############################
2009-12-18 19:52:40,709 WARN
org.apache.zookeeper.server.quorum.QuorumCnxManager: Cannot open
channel to 1 at election address hslave1.test.co.jp/192.168.1.25:3888
java.net.NoRouteToHostException: No route to host
#############################
After all, when I verified the connection status using netstat command,
connection has established among those running 4 server(Master x 1, Slave x
3), and WARN messages would stop.
Regarding this LOG message, I assume it output when a server cannot reach
the target server
during establishing the connetion. Though Log is output at interval of 3
seconds everytime,number of log is not constant(vary from 3 to 8 times). (I
wonder if number of retry may not fixed in each boot?)
I refered "$HBASE_HOME/conf/hbase-site.xml", and changed parameter of
[hbase.zookeeper.property.tickTime] to be 5000 as a trial, but nothing have
changed.
Though I tried to change some other parameter, interval and number of log
output have not changed.
Is there any configuration file which can configure like "number of retry",
"interval of retry",
or "time-out duration" in hbase conf directory? I hope someone could help
me.
My best regards.
--
View this message in context: http://old.nabble.com/zookeeper-setting-tp26950949p26950949.html
Sent from the HBase User mailing list archive at Nabble.com.
Re: zookeeper setting
Posted by Flavio Junqueira <fp...@yahoo-inc.com>.
Hi Hiroshi, The log message you refer to tells you that a server
cannot connect to another one during a leader election step, and it
should not output such messages if the server is following a leader or
if it is the leader. For the meaning of configuration parameters such
as tickTime, you may consider looking at the ZooKeeper documentation:
http://hadoop.apache.org/zookeeper/docs/r3.2.2/zookeeperAdmin.html
Thanks,
-Flavio
On Dec 29, 2009, at 5:03 AM, hiroshi wrote:
>
> Hi,
> I built hadoop(0.20.1)+hbase(0.20.1) cluster structure(Master x 1,
> Slave x
> 4).
> And I tried following process to verify the behavior during system
> failure.Shut down one of the slave server, and start Hbase while I
> have a
> machine which cannot establish connection in the cluster.
> I this regard, I get the following log repeatedly(3-8times)at
> interval of 3
> seconds on "hbase-hadoop-zookeeper-hmaster.log", as a result.
>
> #############################
> 2009-12-18 19:52:40,709 WARN
> org.apache.zookeeper.server.quorum.QuorumCnxManager: Cannot open
> channel to 1 at election address hslave1.test.co.jp/192.168.1.25:3888
> java.net.NoRouteToHostException: No route to host
> #############################
>
> After all, when I verified the connection status using netstat
> command,
> connection has established among those running 4 server(Master x 1,
> Slave x
> 3), and WARN messages would stop.
>
> Regarding this LOG message, I assume it output when a server cannot
> reach
> the target server
> during establishing the connetion. Though Log is output at interval
> of 3
> seconds everytime,number of log is not constant(vary from 3 to 8
> times). (I
> wonder if number of retry may not fixed in each boot?)
>
> I refered "$HBASE_HOME/conf/hbase-site.xml", and changed parameter of
> [hbase.zookeeper.property.tickTime] to be 5000 as a trial, but
> nothing have
> changed.
> Though I tried to change some other parameter, interval and number
> of log
> output have not changed.
>
> Is there any configuration file which can configure like "number of
> retry",
> "interval of retry",
> or "time-out duration" in hbase conf directory? I hope someone could
> help
> me.
>
> My best regards.
> --
> View this message in context: http://old.nabble.com/zookeeper-setting-tp26950949p26950949.html
> Sent from the HBase User mailing list archive at Nabble.com.
>