Hi,

I've got a report of a user of my rpms running Pound 2.4 on RedHat 7.3.
The logs show:

Mar 25 14:51:11 host1 pound: starting...
Mar 25 14:51:15 host1 pound: create thr_resurect: Interrupted system
call - aborted
Mar 25 14:51:15 host1 pound: MONITOR: worker exited normally 1,
restarting...
Mar 25 14:51:15 host1 pound: create thr_resurect: Interrupted system
call - aborted
Mar 25 14:51:15 host1 pound: MONITOR: worker exited normally 1,
restarting...

The patch below seems to fix things for him:

-----------------------
# cat pound24-pthreadfix.patch
--- pound.c.orig        Mon Feb 11 12:53:52 2008
+++ pound.c     Tue Mar 25 15:56:40 2008
(at)(at) -356,7 +356,9 (at)(at)
             /* thread stuff */
             pthread_attr_init(&attr);
             pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
+            /*
             pthread_attr_setschedpolicy(&attr, SCHED_RR);
+            */

 #ifdef  NEED_STACK
             /* set new stack size - necessary for OpenBSD/FreeBSD and
Linux NPTL */

-----------------------

The system is running kernel 2.4.20 and glibc 2.2.5. Any ideas what could
be the proper solution here?

Simon

This is not just a problem on older Linux dists.  Our proxy  
servers run FreeBSD 6.2-RELEASE-p8 and Pound 2.3.2 very successfully  
right now but I have found this morning in testing that I see the  
exact same threading related error with 2.4 on FreeBSD 7.0-RELEASE.   
Making the suggested edit to pound.c clears up the constant death and  
respawn but at this point I've not done a real load test to check for  
negative aspects.[...]

I am running pound 2.4 on a Centos 4 machine with kernel version
(2.6.9-42.0.10.ELsmp RH 3.4.6-3). I don't see that thread error occurring
and we're under a pretty good and steady load. The only errors that I see in
the log are:

Mar 23 19:35:56 pound: (7236c003) error copy server cont: Connection reset
by peer
Mar 23 23:06:41 pound: (72670005) error copy chunk cont: Connection reset by
peer
Mar 24 02:05:00 pound: (72b04003) error read from X.Y.Z.W: Connection timed
out
Mar 24 10:47:24 pound: (73b84003) copy_chunks flush error: Connection reset
by peer
Mar 25 05:52:33 pound: (76f10003) error flush headers to X.Y.Z.W: Connection
reset by peer
[...]

> I am running pound 2.4 on a Centos 4 machine with kernel version[...]

That's right, I'm also running it on RHEL4 without any problem. But on
RedHat 7.3, it doesn't work and as we now found out there are other OS's
where it also fails.

Simon
[...][...][...]

On Wed, 2008-03-26 at 15:51 +0100, Simon Matter wrote:[...]

This is the proper solution - real-time scheduling is nice to have, but
not essential. Given the problem reports posted, we'll just remove it
from future versions.[...]