At 03:36 PM 9/27/99 +0200, Olivier Deckmyn wrote:
I've had exactly the same pb... (I use FreeBSD 3.1) The only solution I found was to increase MAXUSERS...
Thanks Olivier. At least I'm not imagining things. Does this hint at scalability problems if I'm suffering this problem with zero load yet didn't suffer from this on a CGI-based web server with very heavy load ? Or is this (as I hope) just an initial cost of running PCGI/Zope that won't be extrapolated with load ? At 01:42 PM 9/28/99 -0400, Thaxter, Jason M. wrote:
I have had the same setup without having this problem. Python is something of a cpu hog, and I have managed (during load testing) to overload it once or twice (kernel message: Python exited on signal 11) causing a need to restart Zope.
I've had a couple of "Python exit on signal 11" problems too - again on a development machine (ie. next to zero load) but assumed that's because I was doing something spazzy with a 500,000-record DB.
Once I even bounced the box. But once I started running Zope with more threads (8), I was no longer able to kill anything, even when it got swamped.
Ah, ok - maybe I'll try that though I'm almost certain that that will cause more of the first problems (above).
Could you have some issue with disk space or inodes on one of your relevant filesystems?
I don't think so. It's a very well endowed machine <g> and was working completely fine on a diet of pure 100% CGI. chas
----- Message d'origine ----- De : chas <panda@skinnyhippo.com> À : <zope@zope.org> Envoyé : mardi 28 septembre 1999 11:17 Objet : [Zope] FreeBSD : anyone else noticed kernel "table full" errors whenZope is installed ? Zope is installed ?
I have 4 almost identical servers : - FreeBSD 3.2 - MySQL - Apache 1.3.6
All were running various websites using CGI scripts. All were stable and have more than sufficient resources.
I've since installed Zope on 3 of them and all 3 in question have subsequently started throwing up the following kernel log messages as soon as Zope is installed (running on the backend of Apache) :
file: table is full file: table is full file: table is full file: table is full file: table is full
This is easily resolved by increasing the MAXUSERS in the kernel but what worries me is that I should even need to do this. One of the boxes was handling 2 million hits per day last month without creating these kernel errors.
I am tempted to assume that this is all due to Zope because the correlation is so strong - run Zope and these errors occur, don't run zope and don't get any. Even for one of the servers that is still offline with zero traffic.
Has anyone else experienced this on FreeBSD ? Any reason why this might be Zope-induced ?
cheers,
chas