Re: how to identify rogue consumer
Yes. Thanks, Jun On Wed, Oct 8, 2014 at 10:53 PM, Steven Wu wrote: > Jun, you mean trace level logging for requestAppender? > log4j.logger.kafka.network.Processor=TRACE, requestAppender > > if it happens again, I can try to enable it. > > On Wed, Oct 8, 2014 at 9:54 PM, Jun Rao wrote: > > > If enabled request logging, you can find this out. > > > > Thanks, > > > > Jun > > > > On Wed, Oct 8, 2014 at 8:57 PM, Steven Wu wrote: > > > > > I have seen very high "Fetch-Consumer-RequestsPerSec" (like 15K) per > > broker > > > in a relatively idle cluster. My hypothesis some misbehaving consumer > > has a > > > tight polling loop without any back-off logic with empty fetch. > > > > > > Unfortunately, this metric doesn't have per-topic breakdown like > > > "BytesInPerSec" or "MessagesInPerSec". So I can't really tell which > > > topic/consumer is pounding the cluster. > > > > > > Also the storm already ended. So I can't use tcpdump to capture live > > > traffic. > > > > > > So any suggestion? > > > > > >
Re: how to identify rogue consumer
Jun, you mean trace level logging for requestAppender? log4j.logger.kafka.network.Processor=TRACE, requestAppender if it happens again, I can try to enable it. On Wed, Oct 8, 2014 at 9:54 PM, Jun Rao wrote: > If enabled request logging, you can find this out. > > Thanks, > > Jun > > On Wed, Oct 8, 2014 at 8:57 PM, Steven Wu wrote: > > > I have seen very high "Fetch-Consumer-RequestsPerSec" (like 15K) per > broker > > in a relatively idle cluster. My hypothesis some misbehaving consumer > has a > > tight polling loop without any back-off logic with empty fetch. > > > > Unfortunately, this metric doesn't have per-topic breakdown like > > "BytesInPerSec" or "MessagesInPerSec". So I can't really tell which > > topic/consumer is pounding the cluster. > > > > Also the storm already ended. So I can't use tcpdump to capture live > > traffic. > > > > So any suggestion? > > >
Re: how to identify rogue consumer
If enabled request logging, you can find this out. Thanks, Jun On Wed, Oct 8, 2014 at 8:57 PM, Steven Wu wrote: > I have seen very high "Fetch-Consumer-RequestsPerSec" (like 15K) per broker > in a relatively idle cluster. My hypothesis some misbehaving consumer has a > tight polling loop without any back-off logic with empty fetch. > > Unfortunately, this metric doesn't have per-topic breakdown like > "BytesInPerSec" or "MessagesInPerSec". So I can't really tell which > topic/consumer is pounding the cluster. > > Also the storm already ended. So I can't use tcpdump to capture live > traffic. > > So any suggestion? >