The server closes the connection, but the client can't detect and hang about go-nsq HOT 4 CLOSED

All reads and writes in go-nsq are deadlined. You're saying adjusting heartbeat_interval and read_timeout did not result in the client detecting errors and exiting?

It's not clear from your client logs, they only show a window on one second of logs.

from go-nsq.

Thanks for your help. After a little more observation, i find probably this is my handler.HandleMessage issue. Because the go-nsq is blocked at this line:

err := handler.HandleMessage(message)

And my handler's HandleMessage has a operation that sends a processed result to a channel which blocks. I am not sure whether this explains the problem.

Will go-nsq choke if the user has some bad blocking code in handler.HandleMessage? And Is the user's full responsibility to make sure this won't happen or is there already something done is go-nsq to bound this?

from go-nsq.

BTW, there is no more log after the last line of go-nsq log:

{"@timestamp":"2015-08-13T18:30:12+08:00","@version":1,"level":"debug","message":"NSQL \u003e DBG    4 [connector.1/default] (mysite:4150) skip sending RDY 2000 (674 remain out of last RDY 2000)"}

just like the go-nsq is dead. It doesn't do any other operation like reconnect at all.

from go-nsq.

Yes, handlers must be able to make progress for go-nsq to behave correctly because of the way it explicitly manages concurrency.

Going to close this, if after addressing the blocking handler it still doesn't reconnect please re-open.

Thanks!

from go-nsq.