>I am monitoring squid with snmp/mrtg, it helps a lot >when looking at
perfomance issues.
I wasn't aware MRTG could monitor squid.. I'll have to look into that.
>Hmmm, have you tried if the problem persists with a >2.4.x-Kernel? I am
a
>bit conservative when it comes to "new & improved" >releases.
I have not tried the 2.4.x kernel but we did have this same setup in
place and without the utilization spikes on Redhat 9 (2.4 kernel) and
Mandrake 9.1 (2.4 kernel).
>Recommendation by this list for linux is aufs as I >learned a couple of
>day ago. Use reiserfs for the cache-partitions, but >you may be using
>reiserfs already if you are on suse...
I have built and reinstalled from source using aufs. I still get the
utilization spikes, but it seems to have settled a little now it is when
there are a pretty significant amount of users accessing the proxy
concurrently. It still has the same behavior though. X amount of time
very low utilization (2-10 percent cpu) X/2 amoutn of time high
utilization 99 percent solid. I have tried switching to heap
replacement policiies. If this was a 2.6 kernel issue wouldn't other
people also see it?
Ryan
