httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Ames <grega...@remulak.net>
Subject Re: 2.0.23?
Date Wed, 01 Aug 2001 20:56:56 GMT
"Paul J. Reder" wrote:
> 
> Ok, I have run a cvs head build of 2.0.23 (as of Tuesday at 4:30) for 2.75
> hours with the following config and results. The synopsis is that it
> ran well until it died.

hope I do the same :-)

> Config:
>   Server: (with a mirror of apache.org stored locally)
>      Linux box with 384 MB of memory and 30 GB HD running threaded mpm with
>         the apache.org httpd.conf with the following mpm config values.
>         StartServers        3
>         MaxClients          10
>         MinSpareThreads     150
>         MaxSpareThreads     300
>         ThreadsPerChild     50
>         MaxRequestsPerChild 3000
>      Built with a copy of the config.nice from apache.org.

coool!   daedalus has at least 1G of ram  (running prefork on FreeBSD of
course) 

> At about the 2.75 hour mark...
> ==================================================
> Current Time: Wednesday, 01-Aug-2001 10:48:54 EDT
> Restart Time: Wednesday, 01-Aug-2001 08:09:19 EDT
> Parent Server Generation: 0
> Server uptime: 2 hours 39 minutes 35 seconds
> Total accesses: 2259384 - Total Traffic: 63.1 GB
> CPU Usage: u31.79 s96.78 cu0 cs.06 - 1.34% CPU load
> 236 requests/sec - 6.8 MB/second - 29.3 kB/request

from daedalus, a few seconds ago:

46.8 requests/sec - 1.4 MB/second - 30.9 kB/request

So you're driving it about 5x harder than real life, on a box with less
than half the memory.  waaay cool!  Maybe this multithreaded MPM stuff
is a good idea after all ;-)

(of course, daedalus is doing other things besides serving http)

> The interesting portion of the error_log:

> [Wed Aug 01 10:07:59 2001] [notice] child pid 3102 exit signal Segmentation fault (11)
> [Wed Aug 01 10:46:39 2001] [alert] (12)Cannot allocate memory: apr_thread_create:
>                                        unable to create worker thread
> [Wed Aug 01 10:46:39 2001] [alert] (12)Cannot allocate memory: apr_thread_create:
>                                        unable to create worker thread
> [Wed Aug 01 10:46:39 2001] [alert] (12)Cannot allocate memory: apr_thread_create:
>                                        unable to create worker thread
> [Wed Aug 01 10:46:49 2001] [alert] Child 19225 returned a Fatal error...
> 
> At this point Apache stopped serving pages (duh). There was no core file (since
> Apache didn't dump it just exited). There were no core files from the segfaults either.
> I have received core files recently, but I can't pin my finger on why I do or
> don't get them.

I haven't gotten _any_ core dumps with threaded on Linux for ages; wish
I knew the answer.  

btw, I assume you saw the patch Jeff just commited for a memory leak
with sockaddr stuff?  don't know if it's applicable to you, but it might
be worth a try.   

Does anyone know how to force a core dump on Linux? something like gcore
on FreeBSD?  If we had a dump of a fat process (and Jeff's commit
doesn't take care of the memory leak), you could use the "x" command in
gdb to probe into the high end of the heap address range (the part that
doesn't exist in a skinny process), and see if you see any repeating
patterns.

Thanks,
Greg

Mime
View raw message