For this class of problem, DeadlockDebugger should come to the rescue. Florent Gaël Le Mignot wrote:
Hello Zope users !
We are deploying a Zope on a customer's server, but we have regular freeze (Zope stop to answer any requests). The Zope is running Plone 2.1 with some custom products.
Here are all the informations I gathered on the issue:
- Zope version 2.8.4 - Plone version 2.1.1 - Python version 2.3.5 - OS: Debian GNU/Linux, sarge, 2.6.8-smp kernel - Computer: dual Xeon with 4Gb of memory - Zope is behind Apache configured as reversed proxy
Symptoms:
- The Zope stops to answer incoming http request on regular intervals
- When this happens, the CPU isn't used mch (about 2% used), and the memory isn't full (1Gb used on the 4Gb total)
- Inspecting the Python process with gdb shows a normal operation: a running thread, and others waiting in, and then after a while (dettaching gdb, waiting a bit, and reattaching gdb) another thread running. The other threads are in:
#0 0x40026d0b in sem_wait@GLIBC_2.0 () from /lib/tls/libpthread.so.0 #1 0x00000004 in ?? () #2 0x080d29de in PyThread_acquire_lock (lock=0x87b1e70, waitflag=1) at thread_pthread.h:406
- Looking at netstat shows a few established connections to the Zope process, but not much (2-3 on avergae)
What we tried:
- We tried using ZEO, didn't change anything
- We tried with the default 4 threads, with 10 threads, and with up to 40 threads, nothing changed
I'm currently out of ideas to investigate this issue. Does any of you have an idea of why could this happen ? Or an idea of what I could try to undersatnd what's happening ? Any idea is dearly welcomed.
Thanks a lot !
-- Florent Guillaume, Nuxeo (Paris, France) CTO, Director of R&D +33 1 40 33 71 59 http://nuxeo.com fg@nuxeo.com