Squid 1.2 eating CPU time

From: Bill Wichers <billw@dont-contact.us>
Date: Tue, 16 Jun 1998 23:50:40 -0400 (EDT)

I'm having a curious little problem with Squid 1.2.beta22... Nobody seems
to have mentioned this on the list before, so here goes...

I have Squid 1.2.beta22 up and running on two test machines. One is not
loaded (it's a backup box) and the other is very lightly loaded (about 8
connections/minute right now, probably peaking at around 5-6 times that).
Everything seems to be working fine. Pages come up quick, no wierd erros
are being reported to the users or the logs. What is wierd here is that
Squid is eating LOADS and LOADS of CPU time (99.9% according to top), and
appears to be doing some serious thinking about something in the select()
loop since I see the following in cachemgr:

        Select loop called: 1271321570 times, 0.415 ms avg

I think this is probably a bit excessive since Squid has only had about
40k total connections (TCP and ICP included) and has been up a little shy
of 6 days now. The backup machine is using about 48% of the CPU and also
appears to be thinking about the select() loop although not nearly as much
as the lightly loaded machine. On the lightly loaded machine cachemgr

Resource usage for squid:
        UP Time: 527529.577 seconds
        CPU Time: 197501.400 seconds
        CPU Usage: 37.44%
        Maximum Resident Size: 0 KB
        Page faults with physical i/o: 8742

Yeah, I know it's swapping a bit much, but that hasn't caused a problem
like this before. Squid is using about 11 MB of RAM as reported by
mallinfo in cachemgr.

Any ideas? Could this be a bug in 1.2.beta22 or is there an elusive config
option that I should be playing with? Any suggestions much appreciated!

<being thankfull that my cache box is dedicated to Squid>
Received on Tue Jun 16 1998 - 20:51:48 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:43 MST