Date: Sun, 21 Dec 2003 14:35:07 -0800 From: paul beard <paulbeard@mac.com> To: FreeBSD-questions <questions@Freebsd.org> Subject: kern.maxfiles limit exceeded: what to investigate? Message-ID: <EDF26C36-3405-11D8-BA03-000A95BBCCF8@mac.com>
next in thread | raw e-mail | index | archive | help
I upgraded my webserver hardware from a old Pii 233 to an AMD Athlon 700 a few weeks back and seem to be having some teething troubles with it. I have hit the kern.maxfiles limits twice recently, having run for a couple of years without even knowing there was one. I found 420 of these in messages: Dec 21 13:39:30 red /kernel: kern.maxfiles limit exceeded by uid 80, please see tuning(7). and I could only login on console. I read thru tuning(7) and the best I could figure is that my best option was to let the system work its limits based on hardware: accordingly, I set maxusers in my config to 0. I didn't understand the rest of it and I'm not sure much of it applies, since it's not exactly loaded (less than 10000 hits/day). This is what I see now. It looks like a lot of headroom, I think. kern.maxfiles: 4040 kern.maxfilesperproc: 3636 kern.openfiles: 300 Since UID 80 would be the httpd process, I suppose looking into process-specific resource issues is next. I am running Apache 1.3.29. One thing I noticed that seemed a little odd was the snmpd seemed to behaving strangely. I did an snmpwalk to see if I could monitor these kernel values that way (there is so much useful stuff exposed thru snmp), but I found that I couldn't run it more than once, and that snmpd was running at 97% or so of CPU. I have deinstalled and rebuilt it, and now it seems to be behaving properly, but I wonder if every 5 minute snmp requests, in and outbound, with a flaky binary were slowly eating up file descriptors. -- Paul Beard <www.paulbeard.org/> paulbeard [at] mac.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?EDF26C36-3405-11D8-BA03-000A95BBCCF8>