Once the socket bugs are fixed, there will be at least one .-D
Hey, so far everything seems to look like a Swazoo bug, not GST. :-)
I wonder why nobody complains over in AIDA-land, where Swazoo
is used as basis, too. Am I the only one stress-testing his
web app? Not that I would call 25 users a real stress test.
Well, 25 simultaneous cold-cache accesses are pretty heavy, especially
if you consider that most production setups will use a load balancer.
The root cause of the CLOSE_WAIT bug is definitely there in upstream
Swazoo. You had 1000 open sockets, presumably because of keep-alive
behavior of Swazoo (I'll research what apache or lighttpd do) which is
of course nothing that GST cares about---it would be different if say
you were missing close system calls, but nothing seems to hint at a bug
like that. Third, the fact that the accept error message repeats over
and over is there in upstream Swazoo too, and it's trivial to show it:
if you don't have some kind of exponential back-off after an EMFILE
error, you're going to get hte same error over and over again.
But this is very comforting for me, because it shows we're doing things
seriously. Thanks.
Paolo
_______________________________________________
help-smalltalk mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-smalltalk
