apr-bugs mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From bugzi...@apache.org
Subject [Bug 51065] expose socket timeouts in apr_memcache
Date Fri, 19 Oct 2012 13:34:06 GMT
https://issues.apache.org/bugzilla/show_bug.cgi?id=51065

--- Comment #2 from Joshua Marantz <jmarantz@google.com> ---
I patched this change into my development environment and attempted to set the
socket timeout to 10 seconds.  After I did this, the system behaved very poorly
under load, timing out.  Ultimately in my system-load tests, with the socket
timeout set at -1 I was getting >8k queries per second.  But when I set the
socket timeout to 10 seconds I got less than 300 queries per second, with lots
of timeouts.  This load-test was encompassing much more than memcached so the
absolute numbers are not that important, but the dramatic drop is.

My load-test was not expected to induce large numbers of memcached timeouts.  
But the system was much more stable & predictable with the default timeeout of
-1.

Jeff Trawick (trawick@gmail.com) suggested on a thread in dev@apr.apache.org
that this patch "is apparently missing a call to adjust the
socket timeout and to discard the connection if the timeout is
reached", and that seems consistent with my experiments.

So I think some iteration is required before this patch is applied.

-- 
You are receiving this mail because:
You are the assignee for the bug.

---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe@apr.apache.org
For additional commands, e-mail: bugs-help@apr.apache.org


Mime
View raw message