Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
Kafka >> mail # user >> Consumers constantly rebalancing


+
Manish Khettry 2013-02-06, 01:41
+
Neha Narkhede 2013-02-06, 01:47
Copy link to this message
-
Re: Consumers constantly rebalancing
The easiest way to diagnose is to enable GC logging on both the consumer
and the zk instance and see if you have long pauses.

-Jay
On Tue, Feb 5, 2013 at 5:46 PM, Neha Narkhede <[EMAIL PROTECTED]>wrote:

> >> Unable to reconnect to ZooKeeper service, session 0x33c981ab95100ed
> has expired, closing socket connection
>
> This can happen either due to long GC pauses on your client side or due to
> IO pauses on the zookeeper server side.
> That is the reason increasing the session timeout seems to have helped.
> If this error happens frequently, it will cause your consumer instances to
> keep rebalancing.
>
> Thanks,
> Neha
>
>
> On Tue, Feb 5, 2013 at 5:41 PM, Manish Khettry <[EMAIL PROTECTED]> wrote:
>
> > We are trying to trouble shoot a problem wherein our system just cannot
> > seem to read messages fast enough from Kafka. We are on kafka 0.6 and are
> > using the simple consumer.
> >
> > From looking at the logs, and we see a lot (almost constant chatty
> > messages) about rebalancing. So for instance every minute, we see
> messages
> > like this:
> >
> >
> > Consumer rookery-vacuum-prod_<first_ip>.internal-1360106018385
> > rebalancing the following partitions: List(0-0, 0-1, 0-10, 0-11, 0-12,
> > 0-13, 0-14, 0-15, 0-16, 0-17, 0-18, 0-19, 0-2, 0-3, 0-4, 0-5, 0-6,
> > 0-7, 0-8, 0-9, 1-0, 1-1, 1-10, 1-11, 1-12, 1-13, 1-14, 1-15, 1-16,
> > 1-17, 1-18, 1-19, 1-2, 1-3, 1-4, 1-5, 1-6, 1-7, 1-8, 1-9) for topic
> > compact-player-logs with consumers:
> >
> >
> > I also see zookeeper timeouts like so:
> >
> > Unable to reconnect to ZooKeeper service, session 0x33c981ab95100ed
> > has expired, closing socket connection
> >
> >
> > We increased the zookeeper session timeout from 6 seconds to 12 seconds
> and
> > this seems to have helped somewhat but I'm not sure  if these zookeeper
> > timeouts at 6 seconds are symptomatic of a problem with our zookeeper
> > cluster and/or connectivity between the consumers and zk. Any thoughts?
> >
> > Manish
> >
>

 
+
Manish Khettry 2013-02-06, 05:04