Gwen Shapira created KAFKA-3299:
-----------------------------------
Summary: KafkaConnect: DistributedHerder shouldn't wait forever to
read configs after rebalance
Key: KAFKA-3299
URL: https://issues.apache.org/jira/browse/KAFKA-3299
Project: Kafka
Issue Type: Bug
Reporter: Gwen Shapira
Right now, the handleRebalance code calls readConfigToEnd with timeout of
MAX_INT if it isn't the leader.
The normal workerSyncTimeoutMs is probably sufficient.
At least this allows a worker to time-out, get back to the tick() loop and
check the "stopping" flag to see if it should shut down, to prevent it from
hanging forever.
It doesn't resolve the question of what we should do with a worker that
repeatedly fails to read configuration.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
