On 4/16/13 8:24 AM, Vidar Skaugen wrote:
> Hi. We have a web app that uses quite a bit of memory, so after roughly
> 20 instances per GB allocated memory it starts to spend more and more
> time doing garbage collection. With enough concurrent connections it
> grinds to a halt.
> I've been trying to limit the number of threads that are started at
> once. <port-thread-max> seems to work for this, and does what I want
> when I setup Resin as a http server. However I've been unable to get it
> to work with mod_caucho. Setting <port-thread-max> in <port-default> or
> setting port_thread_max in resin.properties does not seem to affect the
> traffic coming in from mod_caucho.
If this is for Resin 4.0, you might try <cluster-port>, which has the
same syntax as <http> and lets you configure the 6800 port properties.
The config looks like:
If you're in a clustered environment, that limit might cause problems
with the clustering.
> The only way I've been able to limit it at all is to set thread-max in
> <server> really low (20-ish seems to be the minimum that has a chance of
> working). Of course I don't want to do this, since this limits all
> system threads and setting it too low causes weird problems if resin
> runs out of threads.
> Setting keepalive-max, connection-max, or thread-idle-max in server
> doesn't help, and accept-thread-max isn't applicable. Is there a setting
> I'm overlooking somewhere?
> - Vidar Engh Skaugen
> resin-interest mailing list
resin-interest mailing list