If your redirectors are CPU bound then just adding more of them won't
help as the CPU is your bottleneck..
Having very many redirectors is a good thing for redirectors which may
wait for external data.
Regards
Henrik
fre 2003-02-21 klockan 19.12 skrev sean.upton@uniontrib.com:
> You are running on a newer SMP Apple machine, right? Have you considered
> just bumping up the number of processes to the compiled max (32?) to ensure
> that you don't run into this problem. If you have a high load-average, you
> likely have one of two problems (or both):
>
> - Too little horsepower or a uniprocessor box.
> - Not enough memory & much CPU usage is from excessive swapping.
>
> There are a lot of people who say you don't need an SMP box for Squid, but
> this is not true in the case of a CPU-bound redirector load. How much
> system memory do you have and how much are your redirectors using? Have you
> considered monitoring load average and memory pressure with something like
> MRTG or another graphing program to figure out where your system's breaking
> point is?
>
> Adding more processes obviously means you are also adding more context
> switches, but if Squid is feeding your redirectors faster than they can
> respond, you likely want to keep the number high for functionality's sake
> (but if the performance goes straight to hell, look for more resources).
>
> One other thing I have always thought is that in some cases it would be nice
> if Squid could do selective caching of existing "queries" to the redirector
> process to avoid the communication latency and CPU requirements of the
> redirector process for all accesses (wishful thinking now, but who knows,
> might be possible in the future).
>
> Sean
>
> -----Original Message-----
> From: Jeff Donovan [mailto:jdonovan@beth.k12.pa.us]
> Sent: Friday, February 21, 2003 7:02 AM
> To: squid-users
> Subject: [squid-users] All redirector processes are busy
>
>
> greetings
> I get this message and it says
> "consider increasing the number of redirector processes in your config
> file"
>
> okay,..I've done that. i started out with 3 and i am up to 20,...and i
> still get it every once in a while. i usually ignore it because things
> seem to work fine. however someone on my network is getting delayed.
>
> Is this normal or is there a better way to streamline the child
> processes.
>
> stable 2.5
> OSX 10.2.3 server
> Cache dir 16gb
> cache mem 256 mb
>
> Also the cpu usage is always running above 60%. most of the time it's
> between 80 and 100%. The docs specify that the cpu speed isn't as
> critical as the drive specs. but it looks to me like i could use a
> quicker processor.
> your thoughts
>
> --jeff
-- Henrik Nordstrom <hno@squid-cache.org> MARA Systems AB, SwedenReceived on Fri Feb 21 2003 - 13:33:26 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:13:33 MST