Cool, thanks for your explanation. We are quite new to kafka and zookeeper. just adopt kafka as a part of our project recently.
It sounds like something failed during rebalancing. I don't have access to our log atm. Will let you know once I do. Thanks for pointing a right direction for us. Cheers, Ke On Mon, Aug 27, 2012 at 12:40 AM, Jun Rao <[email protected]> wrote: > "Owner = null" means that a partition is not consumed by any consumer. > Typically, this only happens transiently when consumes rebalance among > themselves. Do you see lots of rebalances in your consumer log? Do you see > any ZK session expiration? > > Thanks, > > Jun > > On Sun, Aug 26, 2012 at 10:27 AM, Ke Ren <[email protected]> wrote: > > > Hi Jun, > > > > yes, we tried that. we also added mbean over jmx to monitor. The thread > is > > healthy. Moreover, if it's something wrong with application, it should > > recover after restarting. However, after restarting application, it > > received few messages and immediately died. We have to restart kafka and > > zookeeper to make it back to normal. I haven't identified whether those > few > > messages received after restarting are new messages from topic or always > > few repeated messages. > > > > do you know what "Owner = null" means and why it happens? If the consumer > > works properly, owner always equals something like: > > > > Owner = xxxservice_ip-10-60-41-167-1345969396925-4966bd1a-0 > > > > If there is no consumer on a group, ConsumerOffsetChecker won't return > > anything. We always get Owner = null when consumer stops working. Any > > ideas? > > > > Thanks, > > > > Ke > > > > On Sun, Aug 26, 2012 at 6:10 PM, Jun Rao <[email protected]> wrote: > > > > > Ke, > > > > > > Have your tried suggestions in item #3 in > > > http://incubator.apache.org/kafka/faq.html ? If you put your consumer > > > thread in a java threadpool, exceptions may be eaten. > > > > > > Thanks, > > > > > > Jun > > > > > > On Sat, Aug 25, 2012 at 1:29 PM, Ke Ren <[email protected]> wrote: > > > > > > > Hi guys, > > > > > > > > I am running a kafka instance and a zookeeper instance on the same > ec2 > > > > instance. I have 2 consumer groups on the same topic. They work fine > > for > > > > few days but one of the consumers keeps stopping after few days. I > > can't > > > > see any exception in my application logs or kafka log. The thread of > > the > > > > stopping consumer is still running but consumer can't receive any > > > message. > > > > Even I restart my application and it's still the same. I have to > > restart > > > > zookeeper and kafka to make it back to normal. > > > > > > > > I used the following tool suggested by kafka docs and found owner is > > null > > > > after consumer stops working. also its consumer offset won't change > any > > > > more. > > > > > > > > bin/kafka-run-class.sh kafka.tools.ConsumerOffsetChecker --group > > mygroup > > > > --zkconnect 127.0.0.1:2181 --topic mytopic > > > > > > > > It's very similar with > > > > > > > > > > > > > > http://mail-archives.apache.org/mod_mbox/incubator-kafka-users/201207.mbox/%[email protected]%3E > > > > > > > > Any ideas about it? > > > > > > > > Thanks > > > > > > > > Ke > > > > > > > > > >
