Mirrormaker2 mirrorSourceConnector stuck using 1 task

[ntrvic]

I have a mm2 sourceConnector that is responsible for a lot of topic-partitions, however specifically for the sourceConnector, it only runs 1 task. I have the following configuration:

      sourceConnector:
        config:
          offset-syncs.topic.replication.factor: 3
          replication.factor: 3
          replication.policy.class: io.strimzi.kafka.connect.mirror.IdentityReplicationPolicy
          sync.topic.acls.enabled: "false"
        tasksMax: 100
...
    topicsPattern: .*

The connect rest call for tasks looks like:

$ curl localhost:8083/connectors/events-dal-%3Edemo-cluster.MirrorSourceConnector/tasks
[{"id":{"connector":"events-dal->demo-cluster.MirrorSourceConnector","task":0},"config":{"target.cluster.status.storage.replication.factor":"-1","connector.class":"org.apache.kafka.connect.mirror.MirrorSourceConnector","offset-syncs.topic.replication.factor":"3","replication.factor":"3","sync.topic.acls.enabled":"false","topics":".*","replication.policy.class":"io.strimzi.kafka.connect.mirror.IdentityReplicationPolicy","target.cluster.config.storage.replication.factor":"-1","source.cluster.alias":"events-dal","groups":".*","source.cluster.bootstrap.servers":"***********:9092","target.cluster.offset.storage.replication.factor":"-1","target.cluster.alias":"demo-cluster","task.class":"org.apache.kafka.connect.mirror.MirrorSourceTask","name":"events-dal->demo-cluster.MirrorSourceConnector","target.cluster.bootstrap.servers":"demo-cluster-kafka-bootstrap.kafka.svc:9096","task.assigned.partitions": < list of all partitions in our cluster >"

I do not see the same problem with the checkpoint connector. Could this be a bug, or am I missing something obvious?

Thank you!

[scholzj]

Strimzi does not generate the tasks. It just updates the connector. So maybe the first thing you can do is to:

• If the KafkaMirrorMaker2 resource status / operator logs don't contain any errors
• If the Connector configuration reflects your tasksMax change

If that is set correctly, it is up to Kafka to decide how many tasks it would create. But the points above should definitely help to clarify where the issue might be.

[scholzj]

Hmm, that is relatively old and I think a lot of fixes when into MM2 since.

@mimaison Ever heard about a bug like this?

[ntrvic]

Update, I not only needed to restart the workers, but first also had issue a restart on the connector as such curl -X POST kc:8083/connectors/myconnector/restart

[scholzj]

That is weird since restarting the workers will anyway restart the connector IMHO.

[mimaison]

Thanks @ntrvic for the report.

I see your connector name contains a URL encoded character (%3E, >) . Could it be related to https://issues.apache.org/jira/browse/KAFKA-9747?

Otherwise this looks like a bug to me too. Did you manage to collect logs? Can you reproduce the issue?

[ntrvic]

Sorry for the late response -- we were able to reproduce the issue again. Odd behavior this time, it required us to restart the connector a few times. No errors in the logs. We will probably need to update our Strimzi deployment, but we will be on 0.25 for a while as we are on Kafka 2.7. Does it sound like a Strimzi MM2 CRD bug to you or KC? note: I did not have debug logs turned on with the strimzi-operator. If it helps I can probably reproduce this again with them turned on.