I am re-sending this thread from my personal account because my company e-mail (visibletechnologies.com) is not receiving replies from yur mailing list.
To recap, I was seeing a steady stream of exceptions both from the broker and a client. Neha recommended setting ulimit. Our machines were at the default of 1024. Setting the limit to 65536 dramatically reduced the broker-side exception spew. However, repeating the tests from yesterday, I am still seeing the following repeated block of log entries on our client: 2013-03-08 18:44:41,063 INFO kafka.consumer.SimpleConsumer: Reconnect due to socket error: java.io.EOFException: Received -1 when reading from channel, socket has likely been closed. at kafka.utils.Utils$.read(Utils.scala:373) at kafka.network.BoundedByteBufferReceive.readFrom(BoundedByteBufferReceive.scala:67) at kafka.network.Receive$class.readCompletely(Transmission.scala:56) at kafka.network.BoundedByteBufferReceive.readCompletely(BoundedByteBufferReceive.scala:29) at kafka.network.BlockingChannel.receive(BlockingChannel.scala:100) at kafka.consumer.SimpleConsumer.liftedTree1$1(SimpleConsumer.scala:124) at kafka.consumer.SimpleConsumer.kafka$consumer$SimpleConsumer$$sendRequest(SimpleConsumer.scala:122) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply$mcV$sp(SimpleConsumer.scala:161) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply(SimpleConsumer.scala:161) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply(SimpleConsumer.scala:161) at kafka.metrics.KafkaTimer.time(KafkaTimer.scala:33) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply$mcV$sp(SimpleConsumer.scala:160) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply(SimpleConsumer.scala:160) at kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply(SimpleConsumer.scala:160) at kafka.metrics.KafkaTimer.time(KafkaTimer.scala:33) at kafka.consumer.SimpleConsumer.fetch(SimpleConsumer.scala:159) at kafka.server.AbstractFetcherThread.doWork(AbstractFetcherThread.scala:93) at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:50) 2013-03-08 18:44:41,391 INFO kafka.consumer.ConsumerFetcherManager: [ConsumerFetcherManager-1362767252595] removing fetcher on topic VTFull-enriched, partition 0 2013-03-08 18:44:41,392 INFO kafka.utils.VerifiableProperties: Verifying properties 2013-03-08 18:44:41,393 INFO kafka.utils.VerifiableProperties: Property broker.list is overridden to 10.10.2.123:9092 2013-03-08 18:44:41,393 INFO kafka.utils.VerifiableProperties: Property clientid is overridden to K-Router 2013-03-08 18:44:41,393 INFO kafka.client.ClientUtils$: Fetching metadata for topic Set(VTFull-enriched) 2013-03-08 18:44:41,393 INFO kafka.producer.SyncProducer: Connected to 10.10.2.123:9092 for producing 2013-03-08 18:44:41,402 INFO kafka.producer.SyncProducer: Disconnecting from 10.10.2.123:9092 2013-03-08 18:44:41,402 INFO kafka.consumer.ConsumerFetcherManager: [ConsumerFetcherManager-1362767252595] adding fetcher on topic VTFull-enriched, partion 0, initOffset 14935 to broker 0 with fetcherId 0 We are currently using the high-level consumer. Jun's question about java 5 bugs: we are using Java 1.6, so I would assume we have that fix. I will, however,folllow up and see if we can confirm whether this bug is biting us.