[squid-users] High CPU

Still looking into the high CPU for squid. Now averages 88-91% of CPU.

Performed a truss -p <process-id>

Noticed this in the server with high CPU:

poll(0xEFFF7C54, 146, 0) = 1
read(60, " G E T h t t p : / / w".., 4095) = 474
poll(0xEFFF7C54, 147, 0) = 1
write(7, " w w w . y e l l o w\n", 11) = 11
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 0
poll(0xEFFF7C54, 147, 0) = 3
read(61, " G E T h t t p : / / w".., 4095) = 575

I did the same on our failover server, built exactly the same and squid
never sends multilpe polls to the same address. This is occurring through
the entire 'truss' output. Could this indicate a CPU problem?

We have stopped and restarted Squid, however, the CPU usage goes from 0 to
90% in a few seconds. The only things we haven't tried yet is a reboot, but
we'd prefer to find out why this is happening.

Any ideas?

I have set Squid to debug, but that doesn't really show anything up.

Cheers
Received on Wed Dec 26 2001 - 20:58:51 MST