Description

Hello, We have been using the latest Kafka client library (1.2.0) with defaults settings. Our typical Kafka topic consumption loop is to read an event and commit it one by one. Recently we have noticed a lot of random Broker: Unknown member exceptions while commiting event offset.

Logs says:

{"Message":"[thrd:GroupCoordinator]: 
GroupCoordinator/3: Timed out HeartbeatRequest in flight (after 10963ms, timeout #0): possibly held back by preceeding OffsetCommitRequest with timeout in 48457ms",
"ClientInstance":"rdkafka#consumer-1","Facility":"REQTMOUT"} 

then this:

{"Message":"[thrd:GroupCoordinator]: 
GroupCoordinator/3: Timed out 1 in-flight, 0 retry-queued, 0 out-queue, 0 partially-sent requests",
"ClientInstance":"rdkafka#consumer-1","Facility":"REQTMOUT"} 

And finally this happens (because of rebalancing)

Broker: Unknown member ---> 
Confluent.Kafka.KafkaException: Broker: Unknown member\n 
at Confluent.Kafka.Impl.SafeKafkaHandle.Commit(IEnumerable`1 offsets)\n at 
Confluent.Kafka.Consumer`2.Commit(ConsumeResult`2 result)

I’m wondering why we see this preceeding OffsetCommitRequest if we just commit offsets one by one sequentially.

Could you please help to figure out what is happening?

How to reproduce

NuGet packages installed: <PackageReference Include="Confluent.Kafka" Version="1.2.0" />

while (true)
{
                consumeResult = _consumer.Consume(500ms);
                if (consumeResult == null)
                {
                    return;
                }
                _consumer.Commit(consumeResult);
}