Apache Out of Memory
I am running an Apache Webserver (1.3.9-14) for our department on a
mostly potato system.
The largest demand on the server comes from a web-based homework system
which uses a cgi interface. We have 600 students using the system
(dual 550MHz Pentium III with 256M RAM)
Two days ago, I received a message from the Apache Error Logs that said:
[Wed Oct 3 20:10:55 2001] [error] server reached MaxClients setting,
consider raising the MaxClients setting
so I raised the limits from 250 to 512 (the compiled max).
QUESTION 1: How can one determine the number of active clients at a
given time? I would like to monitor this.
The next night with the increased MaxClients setting, the CPU load
suddenly shot to 50, effectively closing off most services. The load
went from 1 to 50 and back down in the space of approximately 30
minutes, and while the load was very high, the Apache error logs kept
posting "Out of Memory Messages"
My guess is that this was a real memory allocation problem (i.e. kernel
based). I have only seen such loads about a year ago when we had random
VM problems, but since I compiled a 2.2.19 kernel, virtual memory has
been rock solid.
QUESTION 2: Could the cgi scripts be responsible for this? I am
guessing this is the case, but I have no idea how to gather evidence.
They are in use at many institutions, so things are not so clear. Any
ideas on how to track the source of this problem down would be appreciated.
Thanks
TRS
--
Thomas R. Shemanske
(Mailing Address) (Office/Internet Information)
Department of Mathematics 203 Choate House
6188 Bradley Hall thomas.r.shemanske@dartmouth.edu
Dartmouth College http://www.math.dartmouth.edu/~trs/
Hanover, NH 03755-3551 (603) 646 - 3179
Directions: http://www.math.dartmouth.edu/~trs/choatehouse.html
Office hours: http://www.math.dartmouth.edu/~trs/frontmatter/office.html
Fall Term Office Hours: M,F 3 - 4:30, W 8:30 - 10.
Reply to: