update to pass over unavailable leaders

[ABaldwinHunter]

@codeclimate/review

This update matches https://github.com/bpot/poseidon/pull/88/files

[ABaldwinHunter]

edit: context : goal is to be able to continue handling messages and work when a node is down

[dasch]

@ABaldwinHunter I've implemented something similar to this in https://github.com/zendesk/ruby-kafka. The code is still early beta, but I would love it if you would give it a look.

[adrien-mogenet]

@ABaldwinHunter : Is it supposed to solve the "NotLeaderForPartition" exception?

[jpignata]

@adrien-mogenet Yes, this was a stab at that. Has that been fixed in some other way in the intervening months?

[adrien-mogenet]

Not AFAIK, at least not directly addressed; a reasonable workaround is to use this project instead:

https://github.com/treasure-data/kafka-fluentd-consumer

[dasch]

This issue is fixed by https://github.com/zendesk/ruby-kafka.

[ABaldwinHunter]

@dasch Thanks for your work and interest! Apologies, I missed your comment before.

Excited to take a look.