Thanks Ewan, but I wonder why is it not fixed yet, seems like a very easy fix - make metadata fetch as part of the code executed in Future and have a timeout. Basically there is more then one issue and both are very critical.
Sent from my iPhone > On Aug 3, 2016, at 01:04, Ewen Cheslack-Postava <e...@confluent.io> wrote: > > This is unfortunate, but a known issue. See > https://issues.apache.org/jira/browse/KAFKA-1894 The producer suffers from > a similar issue with its initial metadata fetch on the first send(). > > -Ewen > > On Thu, Jul 28, 2016 at 12:46 PM, Oleg Zhurakousky < > ozhurakou...@hortonworks.com> wrote: > >> Also, reading java docs for KafkaConsumer#poll(timeout) states: >> >> @param timeout The time, in milliseconds, spent waiting in poll if data is >> not available. If 0, returns >> * immediately with any records that are available now. >> Must not be negative. >> >> Yet even setting it to 0 brings no joy. >> >> Cheers >> Oleg >> >> On Jul 28, 2016, at 3:43 PM, Oleg Zhurakousky < >> ozhurakou...@hortonworks.com<mailto:ozhurakou...@hortonworks.com>> wrote: >> >> So I have KafkaConsumer that is deliberately set with server properties >> pointing to non-running broker. >> Doing KafkaConsumer.poll(100) blocks infinitely even though ‘ >> fetch.max.wait.ms’ is set to 1 millisecond. >> >> Basically I am trying to fail when connection is not possible. >> Any idea how to accomplish that? >> >> Cheers >> Oleg > > > -- > Thanks, > Ewen