0

I have complex python program that uses gevent.

I've been having this horrible issue where its timing out on a socket read for several minutes. I retry multiple times during these episodes and only after several minutes does it succeed.

I'm very sure that the guy listening on the socket is still working prefectly because during one of these episodes because I can read off of that socket from another process that's not evented.

It's possible that this is a bug that has been fixed in gevent 1.0 but upgrading is non-trivial and before spending a week or two upgrading I'd like to be more confident that gevent's sockets are indeed my problem.

It would also be non-trivial to change my program to be non-evented.

How can I test my hypothesis that it's a problem in gevent's socket?

guidoism
  • 7,820
  • 8
  • 41
  • 59

1 Answers1

1

If this is a gevent bug related to libevent there is a good chance that it is fixed in 1.0bx since it uses libev. Give a try at gevent-1.0b3 and figure out.

gwik
  • 679
  • 5
  • 9
  • Yeah I started down that track but it's not trivial to upgrade. I decided to put it on hold after a couple hours of attempting the upgrade. I'd like to test the hypothesis that it's a gevent problem before putting in the time to replacing it. – guidoism Jul 31 '12 at 22:45