Hi, We have Zope 2.6.1 (upgraded) running behind Apache 2.0 on a Redhat 8 dual processor machine with Py 2.1.3. Its been live for quite a few days. Now every night at around 3-5 AM the server sort of stops responding. A ps -ax still shows the process running, and a ping to port 8080 on the machine returns a valid response, just that the server doesnt respond and Apache returns a proxy error. Also before that happens the apache logs show that there are not hits for quite some time. Once I restart the zope process everything sworks fine then. Any ideas as to what might be causing that? Has anybody seen that before?/ My start script: #! /bin/sh umask 077 export INSTANCE_HOME="/var/local/zope" export ZSYSLOG="/dev/log" export EVENT_LOG_FILE="/var/log/zope/Zope.log" cd $INSTANCE_HOME /usr/bin/env bin/python $INSTANCE_HOME/z2.py \ -D \ -Z1 \ -C \ -t 20 \ -z /var/local/zope \ -u zope \ -X \ -a 192.168.10.172 \ -P 8000 \ -m - \ -p - \ -f - \ -M /var/log/zope/detailed.txt \ -l /var/log/zope/zserver.txt TIA AM -- ================================================================== Aseem Mohanty Neurobehavioral Systems Inc, 828 San Pablo Ave, Albany, CA 94706 (R) 510 7696011 (M) 510 3014871 (O) 510 5279231 ================================================================== "I saw `cout' being shifted "Hello world" times to the left and stopped right there!!" -- Steve Gonedes ==================================================================
AM wrote at 2003-3-13 11:23 -0800:
We have Zope 2.6.1 (upgraded) running behind Apache 2.0 on a Redhat 8 dual processor machine with Py 2.1.3. Its been live for quite a few days.
Now every night at around 3-5 AM the server sort of stops responding. A ps -ax still shows the process running, and a ping to port 8080 on the machine returns a valid response, just that the server doesnt respond and Apache returns a proxy error. ... Any ideas as to what might be causing that? Has anybody seen that before?/
We have a similar effect: Zope suddenly stops and sometimes even a SIGTERM is unable to kill it (requires SIGKILL). I was once able to provoke this behaviour by pressing the "minimize" button in "Control_Panel --> Database Management --> Flush Cache". Therefore, I expect the problem to be a deadlock somehow related to invalidation (something requiring two looks acquired in different order at different places). I did not yet seriously analyse the problem. Dieter
Dieter Maurer wrote:
I was once able to provoke this behaviour by pressing the "minimize" button in "Control_Panel --> Database Management --> Flush Cache".
Couldnt reproduce that but thats fine.
Therefore, I expect the problem to be a deadlock somehow related to invalidation (something requiring two looks acquired in different order at different places).
Well quite frankly I didnt understand what you said here. However the debug log for zope does log a conflict but only for the background image or the css file or some such thing (mostly)
I did not yet seriously analyse the problem.
Dieter
How would I go about analysing such a problem. Thanks AM -- ================================================================== Aseem Mohanty Neurobehavioral Systems Inc, 828 San Pablo Ave, Albany, CA 94706 (R) 510 7696011 (M) 510 3014871 (O) 510 5279231 ================================================================== "I saw `cout' being shifted "Hello world" times to the left and stopped right there!!" -- Steve Gonedes ==================================================================
participants (2)
-
AM -
Dieter Maurer