Handling stuck threads

[erelsgl]

When a worker thread gets stuck (for example, in an endless loop or a deadlock), it seems that the server keeps spawning new threads to do the same task, which may make the situation worse.

To check this, I created a small server that, when accepting the "HANG" event, just does a for(;;) loop. This is what I see in the log:

INFO 09:53:46 [New I/O worker #6 ]: start
INFO 09:53:46 [New I/O worker #6 ]: looping forever
INFO 09:54:38 [New I/O worker #8 ]: start
INFO 09:54:38 [New I/O worker #8 ]: looping forever
INFO 09:54:58 [New I/O worker #10 ]: start
INFO 09:54:58 [New I/O worker #10 ]: looping forever
INFO 09:55:18 [New I/O worker #12 ]: start
INFO 09:55:18 [New I/O worker #12 ]: looping forever
INFO 09:55:38 [New I/O worker #14 ]: start
INFO 09:55:38 [New I/O worker #14 ]: looping forever
INFO 09:55:58 [New I/O worker #16 ]: start
INFO 09:55:58 [New I/O worker #16 ]: looping forever
...

[mrniko]

Could you propose own solution? What version of netty do you use?

[erelsgl]

I don't quite understand what is the cause of this - does netty automatically generates a new event, in case the previous event did not return?

Where exactly is this done in the socket.io code? I can try to look at it.

I have netty 3.6.0

[mrniko]

I did not faced with such case, think it's somethere in netty. You can take a look at SocketIOPipelineFactory.getPipeline - it is the message handlers chain builder.