Brian Behlendorf wrote: > > Hmm, this "file table full" problem is worrisome. I'm looking at it right > now, but I'm logged in from Paris at $10/hour and don't want to spend too > long on it. I'm looking at it with "lsof -n" as well as looking through > /var/log/messages to determine what might be leaking fd's. The only idea > I have right now is that it looks like each apache child has a separate fd > to each module in /usr/local/apache2_0_28, which adds up to a lot (30 > modules, 450 child procs = 13,500?)
ooops, I see what you mean about an fd per module per child. That bites big time. I'm very surprised to see them. httpd 97082 nobody txt VREG 13,131072 114273 684290 /usr/local/apache2_0_28/modules/mod_include.so httpd 97082 nobody txt VREG 13,131072 97082 684314 /usr/local/apache2_0_28/modules/mod_autoindex.so httpd 97082 nobody txt VREG 13,131072 52172 684328 /usr/local/apache2_0_28/modules/mod_dir.so httpd 97082 nobody txt VREG 13,131072 49046 684316 /usr/local/apache2_0_28/modules/mod_asis.so We're running with MaxClients 600 now, so the worst case is worse than you stated above. However the average will be less (315 last time I checked, and that's high). Since we lost the parent yesterday that may have aggravated the situation since I believe the dead children became zombies. Will Rowe pointed out that there was an Apache 1.3 release very recently which should drive up traffic. Greg
