The failure in continuous-integration/travis-ci/pr is a KakfaServer Unavailable error..

Can you give an example of the error logging you were having trouble with? I had thought the class name was always included. The standard error formatting is:

[Error {errorno}] {__name__}: {msg}

Well it's not..

For example: I get NodeNorReadyError on every fresh boot. I guess connections are created on demand. Thus when initially requests go out for things like GroupCo-ordinator etc, then the client finds no connection https://github.com/dpkp/kafka-python/blob/master/kafka/client_async.py#L272

The error I get is something like this
2016-07-25 18:38:09,028.28.156042099:kafka.coordinator:123145306509312:ERROR:30790:Error sending GroupCoordinatorRequest_v0 to node 2 [1]

With the change it would appear as

2016-07-25 18:38:09,028.28.156042099:kafka.coordinator:123145306509312:ERROR:30790:Error sending GroupCoordinatorRequest_v0 to node 2 [NodeNotReadyError]

Are you using master? That error formatting was updated in #760

i was using 1.2.5 . Will migrate to the 1.2.6 when it comes. thanks!

Can you also please verify about my hypothesis about NodeNotReadyError? That it's benign and it's expected that connections are built on-demand and thus initial errors are expected..

Yes, that is correct.