ruby-kafka

ruby-kafka copied to clipboard

This PR resolves https://github.com/zendesk/ruby-kafka/issues/941.

As the coordinator waits for each member to rejoin when rebalancing the group, the timeout of join request should be greater than rebalance_timeout.

rebalance_timeout_ms : The maximum time in milliseconds that the coordinator will wait for each member to rejoin when rebalancing the group.

cf. https://kafka.apache.org/protocol.html#The_Messages_JoinGroup

Otherwise, each member might send join requests over and over, and as a result, the following problems occur:

• Some partitions are never processed until session_timeout passes
• It takes much time for rebalance that occurs in a short time to finish