Re: [mongrel2] SSL Weirdness
- Loic d'Anterroches
- 2011-06-25 @ 21:08
On 2011-06-24 00:16, Yaroslav Shirokov wrote:
> I've been having strange behavior out of mongrel2 with SSL enabled.
(develop branch at the time of writing +/- a few commits, reports as 1.7.5
> First ran into a problem where the HAProxy health-check to an SSL'd
mongrel2 server would send it into a 100% CPU spike that never ended,
though it was still able to answer requests.
> Now I have the same behavior when the server is serving legitimate
requests over SSL, proxied through HAProxy just for the bind to privilaged
ports with the healthcheck theoretically disabled.
> I've attached an strace snippit of what it looks like when the server is
in the spike. The list of FD's that are being looped through poll, then
being recvdfrom() == 0'd slowly grows over time.
> Anyone have any idea, or even, has anyone else seen this before?
Is the health check against an handler or a Dir? If an handler, are you
handling the disconnect messages correctly (ignoring or updating your
internal state)? I had a spike when my handler was getting from mongrel2
a client disconnect message and I was answering a 404 to the client
which was of course not here anymore. Thus snowballing.
haproxy with ssl may trigger this issue but not your previous setup.
Just an idea, I have not tested yet ssl as I wanted :/
Indefero - Project management and code hosting - http://www.indefero.net
Photon - High Performance PHP Framework - http://photon-project.com
Céondo Ltd - Web + Science = Fun - http://www.ceondo.com