Since I upgraded from RH9 to FC2, I've been having this problem,
roughly twice daily: I'll come home, and my machine is swapping
madly. Logging in from another machine shows something like:
top - 11:07:41 up 11:58, 3 users, load average: 9.64, 9.01, 7.60
Tasks: 134 total, 2 running, 129 sleeping, 0 stopped, 3 zombie
Cpu(s): 20.1% us, 5.6% sy, 3.6% ni, 36.7% id, 33.8% wa, 0.0% hi, 0.2% si
Mem: 1034404k total, 1029672k used, 4732k free, 744k buffers
Swap: 1028152k total, 267844k used, 760308k free, 8456k cached
Now here's the weird thing: I can't tell which process is holding
on to nearly 1GB of memory.
If I kill X, everything settles down and goes back to normal; so
that suggests that the culprit is the X server or some client of
it. But I can't tell which.
Top (and ps, and /dev/*/map) says that the X server, in times like
this, is only around 150MB. It is the largest process. Number two
is Mozilla, at a svelte 89MB.
Killing Mozilla alone does not fix matters. I spent some time
killing off other X clients one at a time, but I wasn't totally
methodical about it, so I didn't find a culprit that way either.
So, clearly some program is going nuts on me here, but the really
weird thing is that I can't tell who all the memory belongs to.
Someone suggested trying "echo 100 > /proc/sys/vm/swappiness" to
see what happens, but that didn't cause any change that I could see.
Any suggestions for what I should look for the next time this
happens? Pretty much guarenteed that it will recur within 12
or 14 hours.
FC2; I've seen this on both kernel-smp-2.6.7-1.494.2.2 and
kernel-smp-2.6.8-1.521. Two "AMD Athlon(tm) MP 2800+" on a
GA-7DPXDW-P mobo.
Thanks,
--
Jamie Zawinski
jwz@xxxxxxx http://www.jwz.org/
jwz@xxxxxxxxxxxxx http://www.dnalounge.com/
