Re: 0.10.1.0 - KafkaConsumer.poll() blocks background heartbeat thread causing consumer to be considered dead?

Hey Jaikiran, Thanks for the report. We shouldn't hold the lock in poll() longer than the heartbeat interval, so if you're seeing this, it's probably a bug. Let me see if I can reproduce it. One quick question about the code snippet. What kind of executor are you using? Is the load on the topic sm

Re: 0.10.1.0 - KafkaConsumer.poll() blocks background heartbeat thread causing consumer to be considered dead?

Thanks Ismael. Just checked, that one doesn't look like it's the same issue, but could be a similar one. In that JIRA it looks like the issue was probably addressed for the commitSync call. However, in this specific instance the KafkaConsumer.poll(...) itself leads to locking the object monitor

Re: 0.10.1.0 - KafkaConsumer.poll() blocks background heartbeat thread causing consumer to be considered dead?

Maybe https://issues.apache.org/jira/browse/KAFKA-4303? On 2 Nov 2016 10:15 am, "Jaikiran Pai" wrote: > We have been trying to narrow down an issue in 0.10.1 of Kafka in our > setups where our consumers are marked as dead very frequently causing > rebalances almost every few seconds. The consume

0.10.1.0 - KafkaConsumer.poll() blocks background heartbeat thread causing consumer to be considered dead?

We have been trying to narrow down an issue in 0.10.1 of Kafka in our setups where our consumers are marked as dead very frequently causing rebalances almost every few seconds. The consumer (Java new API) then starts seeing exceptions like: org.apache.kafka.clients.consumer.CommitFailedExcepti