Home | About | Sematext search-lucene.com search-hadoop.com
NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB
 Search Hadoop and all its subprojects:

Switch to Threaded View
Zookeeper >> mail # user >> zookeeper ephemeral node expired time not accurate


Copy link to this message
-
Re: zookeeper ephemeral node expired time not accurate
See
http://zookeeper.apache.org/doc/r3.2.2/zookeeperAdmin.html#sc_configurationfor
more information.

The minimum session expiration time is a small multiple of this number so
it isn't surprising that you still see inaccuracy.

Have you tried setting it to a very small value like 200?

2011/12/31 朱晨杰 <[EMAIL PROTECTED]>

> Hi, Ted:
>   In the Zookeeper config file, I set tickTime=1000, does this parameter
> decide the frequency of heartbeat? I have change this parameter to
> different value, but the expiration time is still not accurate
>
> 2012/1/1 Ted Dunning <[EMAIL PROTECTED]>
>
> > The client sends heart-beats to the server.  When the time since the last
> > heartbeat is long enough, the server deems the client to have failed or
> > been partitioned away from the cluster.
> >
> > If you have heartbeats every 5s and an session expiration time of 10s,
> then
> > if you kill the connection just before the next heartbeat, the server
> would
> > be justified in expiring the session 5+epsilon seconds after the
> connection
> > is lost.
> >
> > You can configure heartbeats to occur more often if you want.
> >
> > On Sat, Dec 31, 2011 at 4:32 AM, 朱晨杰 <[EMAIL PROTECTED]> wrote:
> >
> > > Hi, everyone:
> > >    I have done some tests and find that the expired time of ephemeral
> > node
> > > is not accurate. I set my connection's timeout to be "10s" when
> > connecting
> > > to zookeeper. Then I create an ephemeral node. I start another
> > application
> > > set watch on this ephemeral node. When I close the first connection,
> the
> > > watch will be triggered and I got the message. I compare the time when
> I
> > > close the connection and the time when I got the message reporting the
> > > disappearance of the ephemeral node, I find it's always not 10 seconds.
> > > Most of the situation, it will be less than 10 seconds, about 8, or 7,
> > even
> > > 6 seconds. Can anyone tell me why this happen? I want to figure out how
> > > does zookeeper decide a connection expire. Thanks
> > >
> > > --
> > > Zhu Chenjie
> > > Zhejiang University, China
> > >
> >
>
>
>
> --
> Zhu Chenjie
> Zhejiang University, China
>
NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB