Sean Gabriel Heacock wrote:
On Tue, 2006-08-29 at 08:27 +0300, Janno Sannik wrote:
Running it on production end. Everything works as expected, but
processors tend to die off and it ends with apache not serving any
requests. restart of apache will fix that.
Processes should die out after a certain (configured) amount of
inactivity, but if more requests come in, new processes are supposed to
be spawned. It shouldn't ever "run out", is that what appears to be
happening?
Sry I ment multiplexors. So the thing is that apache is not serving
requests and service httpd status shows me bunch of pids running.
Before that apache log has 20-30 of these errors:
[Mon Aug 28 23:38:17 2006] [warn] (9)Bad file descriptor:
apr_socket_opt_set: (TCP_NODELAY)
[Mon Aug 28 23:38:17 2006] [warn] (9)Bad file descriptor:
apr_socket_opt_set: (TCP_NODELAY)
[Mon Aug 28 23:38:17 2006] [warn] (9)Bad file descriptor:
apr_socket_opt_set: (TCP_NODELAY)
[Mon Aug 28 23:38:17 2006] [warn] (9)Bad file descriptor:
apr_socket_opt_set: (TCP_NODELAY)
[Mon Aug 28 23:38:18 2006] [warn] (9)Bad file descriptor:
apr_socket_opt_set: (TCP_NODELAY)
I would also like to report some weird thing about this. 3 days ago
peruser "hanged" just like it sometimes does. Did a restart and 2 hours
later same thing happened. Some really buggy client?
It seems that when apache+peruser stops responding, it's usually because
a client is connected to the multiplexer but not issuing a request.
Configuring several multiplexers, via multiple Multiplexer configuration
directives. I configured my setup to have three of them, and since then
apache almost never stops responding like that. It should also make
apache a little more responsive, but I haven't done any benchmarks.
Umm...I have 10 multiplexers :S
So could it be problem with compilation not the peruser itself ?
I would be cool if it is, because i thought it was problem with peruser.
_______________________________________________
Peruser mailing list
[email protected]
http://www.telana.com/mailman/listinfo/peruser
