[
https://issues.apache.org/jira/browse/KAFKA-2568?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Gustafson resolved KAFKA-2568.
------------------------------------
Resolution: Fixed
We ended up fixing the main problem here in KAFKA-2683. The consumer will raise
an exception to the user if it receives an unexpected response from the server.
However, if the server sends no response and just disconnects, the consumer
will still block indefinitely. This will be addressed in KAFKA-2391, so I'm
resolving this.
> New consumer hangs indefinitely when broker list is misconfigured
> -----------------------------------------------------------------
>
> Key: KAFKA-2568
> URL: https://issues.apache.org/jira/browse/KAFKA-2568
> Project: Kafka
> Issue Type: Bug
> Affects Versions: 0.9.0.0
> Reporter: Jason Gustafson
> Assignee: Jason Gustafson
>
> If you accidentally point the broker list configuration to something other
> than an 0.9 Kafka cluster, the consumer will hang indefinitely in poll()
> while it tries to discover the coordinator. This is less than ideal for users
> since it provides no indication of the problem. An incompatible server could
> respond in two ways:
> 1. It may send back an invalid response.
> 2. It may just terminate the connection.
> The first case is straightforward: we should get an ApiException which can be
> propagated to the user. The second case is trickier since it's difficult to
> distinguish this case from a regular broker failure. We might see this happen
> if the new consumer is pointed to an 0.8 broker for example. I'm not sure
> there's much we can do in this case other than making the disconnects visible
> in logging, but that is better than eating the errors silently.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
