You are not logged in Log in Join
You are here: Home » Members » Tres Seaver's Zope.org Site » Various Projects » HighlyAvailableZope » RunningInCircles » wikipage_view

Log in
Name

Password

 
 
FrontPage »

RunningInCircles

Running In Circles

This page is for discussing situations which cause Zope to eat CPU cycles and stop responding.

  • I've seen this happen mostly when updating catalogs. I'm just not sure if it was when I had subtransactions on and SQL in the same transaction (bad), or with subtransactions off (bad - especially when you don't have enough memory). (ItamarST?)
  • Random occurences. No idea why this happens. (ItamarST?)


John Sutherland posted this bit of advice to the Zope list.

I realize that this is not a recommended solution, but this made a HUGE difference to me..

Modify your /usr/lib/python1.5/config/Makefile.pre.in (in Redhat6.2/Mandrake7.0) to have the line (matching the options for your system):

  CFLAGS=         -O6 -mpentium -march=pentium -mcpu=k6 -I$(INCLUDEPY) \
                  -I$(EXECINCLUDEPY) $(DEFS)

Then recompile Zope... This made a rather huge increase in response my system.. Pages accessed locally comeup nearly as fast as the browser can render them (for the management screens anyways..)


Simon Coles noted:
[email protected]
For reasons unknown we lost http connectivity to our Zope server three times last night. I've checked the http logs and, as I suspected, didn't find anything unusual around the time we lost connectivity. The Python processes are still running and everything looks normal except for the connectivity problem. Does anybody have any idea what might be causing this? I haven't found anything.
operating system
Linux 2.2.12-20smp (Redhat 6.1)

We had something similar on a number of Red Hat Linux boxes we run (both single and multiprocessor). The machine would become uncommunicative for a while and then magically come back. No apparent reasons why, no logs, the kind of thing that makes you doubt your sanity.

The solution was to replace the 3Com ethernet card driver with the latest version (I got it from the Linux NIC driver page).

You'll need the kernel source, replace the 3c59x.c file with the latest one. Then like "make modules" and then copy the resulting 3c59x.o file into /lib/modules/ directory). Reboot....

Problem went away for us, machine nice and stable. Don't know the detail around what the problem was or why this solves it.