Rebalance happens due to long running consumer thread [#t1] - Will the same thread join the rebalance?

Question

I want to know what happens to the current processing consumer thread when the rebalance occurs?

Consider three consumers threads are running and I have three partitions for a topic x.

• max.poll.interval.ms= 500000

• session.timeout.ms = 50000

• heartbeat.interval.ms = 15000

• max.poll.records = 1

• consumer thread #t1 -> calls some service and waits till socket timeout which is infinite. That means the thread #t1 will not call poll again in worst case. Once it exceeds the 500000 time this rebalance is triggered

• consumer thread #t2 and consumer thread #t3-> will participate in rebalance on the poll() method call as they were working fine.

• Now, What happens to thread t1 at the time of rebalance?

• Will that thread t1 be killed and will be considered at the time of reassignment of partition

• OR will it keep on running continuously?

Because we see in our application onRebalance all the threads t1, t2 and t3 are getting partition assigned again. Then how and where in code does the thread get killed?

Using

• spring-boot version -- 2.2.6 release
• spring-kafka - 2.3.7 release
• kafka-client -2.3.1
• apache-kafka - kafka_2.12-2.3.1

Answer 1

The thread will not be killed (or interrupted, but socket reads are not interruptible anyway, you need to set a timeout).

When the thread finally returns (if it does), it will trigger another rebalance.