kafka-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Harsha <ka...@harsha.io>
Subject Re: kafka brokers going down within 24 hrs
Date Fri, 16 Jan 2015 19:54:57 GMT
Tousif,
        I meant to say if kafka broker is going down often its better to
        analyze whats the root of cause of the crash.  Using supervisord
        to monitor kafka broker is fine, sorry about the confusion.
-Harsha
On Fri, Jan 16, 2015, at 11:25 AM, Gwen Shapira wrote:
> Those errors are expected - if broker 10.0.0.11 went down, it will
> reset the connection and the other broker will close the socket.
> However, it looks like 10.0.0.11 crashes every two minutes?
> 
> Do you have the logs from 10.0.0.11?
> 
> On Thu, Jan 15, 2015 at 9:51 PM, Tousif <tousif.pasha@gmail.com> wrote:
> > i'm using kafka 2.9.2-0.8.1.1 and zookeeper 3.4.6.
> > i noticed that only one broker is going down.
> >  My message size is less thn 3 kb and  KAFKA_HEAP_OPTS="-Xmx512M"
> > and  KAFKA_JVM_PERFORMANCE_OPTS="-server -XX:+UseCompressedOops
> > -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:+CMSClassUnloadingEnabled
> > -XX:+CMSScavengeBeforeRemark -XX:+DisableExplicitGC
> > -Djava.awt.headless=true" .
> >
> >  Do you mean kafka broker never goes down and  does broker start
> > automatically after failing ?
> > I see only these errors on both the brokers.
> >
> > 10.0.0.11 is the broker which is going down.
> >
> > ERROR Closing socket for /10.0.0.11 because of error
> > (kafka.network.Processor)
> > java.io.IOException: Connection reset by peer
> > at sun.nio.ch.FileDispatcher.read0(Native Method)
> > at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:21)
> > at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:198)
> > at sun.nio.ch.IOUtil.read(IOUtil.java:171)
> > at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:245)
> > at kafka.utils.Utils$.read(Utils.scala:375)
> > at
> > kafka.network.BoundedByteBufferReceive.readFrom(BoundedByteBufferReceive.scala:54)
> > at kafka.network.Processor.read(SocketServer.scala:347)
> > at kafka.network.Processor.run(SocketServer.scala:245)
> > at java.lang.Thread.run(Thread.java:662)
> > [2015-01-16 11:01:48,173] INFO Closing socket connection to /10.0.0.11.
> > (kafka.network.Processor)
> > [2015-01-16 11:03:08,164] ERROR Closing socket for /10.0.0.11 because of
> > error (kafka.network.Processor)
> > java.io.IOException: Connection reset by peer
> > at sun.nio.ch.FileDispatcher.read0(Native Method)
> > at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:21)
> > at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:198)
> > at sun.nio.ch.IOUtil.read(IOUtil.java:171)
> > at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:245)
> > at kafka.utils.Utils$.read(Utils.scala:375)
> > at
> > kafka.network.BoundedByteBufferReceive.readFrom(BoundedByteBufferReceive.scala:54)
> > at kafka.network.Processor.read(SocketServer.scala:347)
> > at kafka.network.Processor.run(SocketServer.scala:245)
> > at java.lang.Thread.run(Thread.java:662)
> > [2015-01-16 11:03:08,280] INFO Closing socket connection to /10.0.0.11.
> > (kafka.network.Processor)
> > [2015-01-16 11:03:48,369] ERROR Closing socket for /10.0.0.11 because of
> > error (kafka.network.Processor)
> > java.io.IOException: Connection reset by peer
> > at sun.nio.ch.FileDispatcher.read0(Native Method)
> > at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:21)
> > at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:198)
> > at sun.nio.ch.IOUtil.read(IOUtil.java:171)
> > at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:245)
> > at kafka.utils.Utils$.read(Utils.scala:375)
> > at
> > kafka.network.BoundedByteBufferReceive.readFrom(BoundedByteBufferReceive.scala:54)
> > at kafka.network.Processor.read(SocketServer.scala:347)
> > at kafka.network.Processor.run(SocketServer.scala:245)
> > at java.lang.Thread.run(Thread.java:662)
> >
> >
> >
> > On Thu, Jan 15, 2015 at 7:49 PM, Harsha <kafka@harsha.io> wrote:
> >
> >> Tousif,
> >>        Which version of kafka and zookeeper are you using and whats your
> >>        message size and jvm size that you allocated for kafka brokers.
> >> There is only 1 zookeeper node , if its a production cluster I recommend
> >> you to have quorum of zookeeper nodes. Both kafka & storm are heavy
> >> users of zookeeper. Also supervisord is recommended for storm I am not
> >> sure you need to have it for kafka, for storm its the fail-fast nature
> >> of workers that requires supervisord to restart.
> >> When kafka goes down first time , i.e before supervisord restarts it do
> >> you see same OOM error. Check the logs to see why its going down for the
> >> first time.
> >> -Harsha
> >>
> >>
> >>
> >> On Wed, Jan 14, 2015, at 10:50 PM, Tousif wrote:
> >> > Hello Chia-Chun Shih,
> >> >
> >> > There are multiple issues,
> >> > First thing is i don't see out of memory error every time and OOM happens
> >> > after supervisord keep retrying to start  kafka.
> >> > It goes down when it tries to add partition fetcher
> >> >
> >> > it starts with
> >> >
> >> > *conflict in /controller data:
> >> > {"version":1,"brokerid":0,"timestamp":"1421296052741"} stored data:
> >> > {"version":1,"brokerid":1,"timestamp":"1421291998088"}
> >> > (kafka.utils.ZkUtils$)*
> >> >
> >> >
> >> > ERROR Conditional update of path
> >> > /brokers/topics/realtimestreaming/partitions/1/state with data
> >> >
> >> {"controller_epoch":34,"leader":0,"version":1,"leader_epoch":54,"isr":[0]}
> >> > and expected version 90 failed due to
> >> > org.apache.zookeeper.KeeperException$BadVersionException: KeeperErrorCode
> >> > =
> >> > BadVersion for /brokers/topics/realtimestreaming/partitions/1/state
> >> > (kafka.utils.ZkUtils$)
> >> >
> >> > and then
> >> >
> >> > [ReplicaFetcherManager on broker 0] Removed fetcher for partitions
> >> > [realtimestreaming,0],[realtimestreaming,1]
> >> > (kafka.server.ReplicaFetcherManager)
> >> > [2015-01-15 09:57:34,350] INFO Truncating log realtimestreaming-0 to
> >> > offset
> >> > 846. (kafka.log.Log)
> >> > [2015-01-15 09:57:34,351] INFO Truncating log realtimestreaming-1 to
> >> > offset
> >> > 957. (kafka.log.Log)
> >> > [2015-01-15 09:57:34,650] INFO [ReplicaFetcherManager on broker 0] *Added
> >> > fetcher for partitions ArrayBuffer([[realtimestreaming,0], initOffset 846
> >> > to broker id:1,host:realtimeslave1.novalocal,port:9092] ,
> >> > [[realtimestreaming,1], initOffset 957 to broker
> >> > id:1,host:realtimeslave1.novalocal,port:9092] )
> >> > (kafka.server.ReplicaFetcherManager)*
> >> > [2015-01-15 09:57:34,654] INFO [ReplicaFetcherThread-0-1], Starting
> >> >  (kafka.server.ReplicaFetcherThread)
> >> > [2015-01-15 09:57:34,747] INFO [ReplicaFetcherThread-1-1], Starting
> >> >  (kafka.server.ReplicaFetcherThread)
> >> > [2015-01-15 09:58:14,156] INFO Closing socket connection to /10.0.0.11.
> >> > (kafka.network.Processor)
> >> >
> >> >
> >> >
> >> > On Thu, Jan 15, 2015 at 12:01 PM, Chia-Chun Shih
> >> > <chiachun.shih@gmail.com>
> >> > wrote:
> >> >
> >> > > You can use tools (e.g., VisialVM) to diagnose OOM problem.
> >> > >
> >> > > 2015-01-15 14:15 GMT+08:00 Tousif Khazi <tousif@senseforth.com>:
> >> > >
> >> > > > i see this error
> >> > > >
> >> > > >  ERROR [ReplicaFetcherThread-0-1], Error for partition
> >> > > > [realtimestreaming,1] to broker 1:class
> >> > > > kafka.common.NotLeaderForPartitionException
> >> > > > (kafka.server.ReplicaFetcherThread)
> >> > > > [2015-01-15 10:00:04,348] INFO [ReplicaFetcherManager on broker
0]
> >> > > > Removed fetcher for partitions [realtimestreaming,1]
> >> > > > (kafka.server.ReplicaFetcherManager)
> >> > > > [2015-01-15 10:00:04,355] INFO Closing socket connection to
> >> > > > /10.0.0.11. (kafka.network.Processor)
> >> > > > [2015-01-15 10:00:04,444] WARN [KafkaApi-0] Fetch request with
> >> > > > correlation id 0 from client ReplicaFetcherThread-0-0 on partition
> >> > > > [realtimestreaming,1] failed due to Leader not local for partition
> >> > > > [realtimestreaming,1] on broker 0 (kafka.server.KafkaApis)
> >> > > > [2015-01-15 10:00:04,545] INFO [ReplicaFetcherThread-0-1], Shutting
> >> > > > down (kafka.server.ReplicaFetcherThread)
> >> > > > [2015-01-15 10:00:04,848] INFO [ReplicaFetcherThread-0-1], Stopped
> >> > > > (kafka.server.ReplicaFetcherThread)
> >> > > > [2015-01-15 10:00:04,849] INFO [ReplicaFetcherThread-0-1], Shutdown
> >> > > > completed (kafka.server.ReplicaFetcherThread)
> >> > > > [2015-01-15 10:00:39,256] ERROR Closing socket for /10.0.0.11
> >> because
> >> > > > of error (kafka.network.Processor)
> >> > > > java.io.IOException: Connection reset by peer
> >> > > > at sun.nio.ch.FileDispatcher.read0(Native Method)
> >> > > > at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:21)
> >> > > > at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:198)
> >> > > > at sun.nio.ch.IOUtil.read(IOUtil.java:171)
> >> > > > at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:245)
> >> > > >
> >> > > > On Wed, Jan 14, 2015 at 10:12 PM, Tousif <tousif.pasha@gmail.com>
> >> wrote:
> >> > > > > Thanks harsha for quick response.
> >> > > > > I don't see any other error. I used to see replica fetcher
error
> >> but
> >> > > > seems
> >> > > > > to be disappeared after setting replica fetcher threads
to 2 as I
> >> have
> >> > > 2
> >> > > > > partitions. Some times I see zookeeper session expiration.
> >> > > > > On Jan 14, 2015 8:02 PM, "Harsha" <kafka@harsha.io>
wrote:
> >> > > > >
> >> > > > >> Tousif,
> >> > > > >>        Do you see any other errors in server.log
> >> > > > >> -Harsha
> >> > > > >>
> >> > > > >> On Wed, Jan 14, 2015, at 01:51 AM, Tousif wrote:
> >> > > > >> > Hello,
> >> > > > >> >
> >> > > > >> > I have configured kafka nodes to run via supervisord
 and see
> >> > > > following
> >> > > > >> > exceptions
> >> > > > >> > and eventually brokers going out of memory. i have
given enough
> >> > > memory
> >> > > > >> > and
> >> > > > >> > process 1 event/second. kafka goes down every day
> >> > > > >> >
> >> > > > >> > i'm wondering what configurastion is missing or
need to be added
> >> > > > >> >
> >> > > > >> > Here are my cluster details:
> >> > > > >> >  2 brokers
> >> > > > >> >  1 zookeeper
> >> > > > >> > and 2 node apache storm
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > INFO zookeeper state changed (SyncConnected)
> >> > > > >> > (org.I0Itec.zkclient.ZkClient)
> >> > > > >> > ERROR Closing socket for /10.0.0.11 because of
error
> >> > > > >> > (kafka.network.Processor)
> >> > > > >> > java.io.IOException: Connection reset by peer
> >> > > > >> > at sun.nio.ch.FileDispatcher.read0(Native Method)
> >> > > > >> > at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:21)
> >> > > > >> > at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:198)
> >> > > > >> > at sun.nio.ch.IOUtil.read(IOUtil.java:171)
> >> > > > >> > at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:245)
> >> > > > >> > at kafka.utils.Utils$.read(Utils.scala:375)
> >> > > > >> > at
> >> > > > >> >
> >> > > > >>
> >> > > >
> >> > >
> >> kafka.network.BoundedByteBufferReceive.readFrom(BoundedByteBufferReceive.scala:54)
> >> > > > >> > at kafka.network.Processor.read(SocketServer.scala:347)
> >> > > > >> > at kafka.network.Processor.run(SocketServer.scala:245)
> >> > > > >> > at java.lang.Thread.run(Thread.java:662)
> >> > > > >> > [2015-01-13 23:43:37,962] INFO Closing socket connection
to /
> >> > > > 10.0.0.11.
> >> > > > >> > (kafka.network.Processor)
> >> > > > >> > Error occurred during initialization of VM
> >> > > > >> > Could not reserve enough space for object heap
> >> > > > >> > Error occurred during initialization of VM
> >> > > > >> > Could not reserve enough space for object heap
> >> > > > >> >
> >> > > > >> >
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > --
> >> > > > >> > Regards,
> >> > > > >> > Tousif
> >> > > > >> > +918050227279
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > --
> >> > > > >> >
> >> > > > >> >
> >> > > > >> > Regards
> >> > > > >> > Tousif Khazi
> >> > > > >>
> >> > > >
> >> > > >
> >> > > >
> >> > > > --
> >> > > > Regards,
> >> > > > Tousif
> >> > > > +918050227279
> >> > > >
> >> > >
> >> >
> >> >
> >> >
> >> > --
> >> >
> >> >
> >> > Regards
> >> > Tousif Khazi
> >>
> >
> >
> >
> > --
> >
> >
> > Regards
> > Tousif Khazi

Mime
View raw message