Enno Runne
Enno Runne
The committer is inherently reducing the incoming data as the elements are collected until the highest offsets get committed. The signature of the flow would be `FlowWithContext[E, Committable, immutable.Seq[E], CommittableOffsetBatch,...
https://travis-ci.org/akka/alpakka-kafka/jobs/625389283#L983 in #1005 https://travis-ci.org/akka/alpakka-kafka/jobs/625389285#L986 in #1005
https://travis-ci.org/akka/alpakka-kafka/jobs/625607991#L978 in #926
I'm trying to understand how this can be a 15s timeout https://travis-ci.org/akka/alpakka-kafka/jobs/625654499#L979-L982 So it must be that the actor is already terminated when the commit is sent.
It is indeed that the first consumer actor is terminated before the commit message is sent to it. This is the scenario where it is important to have a `stop-timeout`...
Experiencing it repeatedly today https://travis-ci.org/akka/alpakka-kafka/jobs/449473247
See even #601 It happened again in https://travis-ci.org/akka/alpakka-kafka/jobs/615474265#L439
https://travis-ci.org/akka/alpakka-kafka/jobs/616606436#L539
Recent failures are different than the original report: ``` [info] partitioned source [info] - should resume topics with demand *** FAILED *** (10 seconds, 347 milliseconds) [info] java.lang.AssertionError: Expected OnNext(_),...
When the Kafka client reports `NotLeaderForPartitionException` the actor polling Kafka will stop. This far it did not pass the cause for stopping to the stream stage. This is now implemented...