Hi,
Thanks for the response Hendrik.
I looked into helper.c and in the Enqueue and
StatefulEnqueue methods, I noticed that a fatal error
is being thrown when the helper stats queue size is
more that twice the number of running redirectors.
Firstly, I want to know if I could change this to
multiples of 3 or 4. Will it affect the functioning
any where else? Secondly, what is the corrective
measure that is being applied by restarting squid in
case of such fatal errors.
Plz help me understand this better.
Regards and TIA,
Deepa
--- Hendrik_Voigtl�nder <hendrik@voigtlaenders.net>
wrote: > Hi
>
> > Cache takes a considreable amt of memory which
> our
> > system cannot afford, hence the choice of disbling
> the
> > cache.
>
> Probably it is a better idea to get one or more of
> cheap intel-based
> machines, 1GB should be affordable with this
> machines.
>
> > The redirector that I am using is a self made
> one
> > for content filtering. It communicates with a
> remote
> > server to obtain the details of the content of the
> > url. This communication alone could take a max of
> > 2000msecs.
>
> We have a servicetimes something like 300ms when
> under load, hit service
> time is about 5ms. Adam Aube told me:
> "Anything under ~ 1 seconds is probably fine for
> misses, and even up to
> 2 seconds depending on congestion and latency on
> your link."
> In your case the servicetime will be more than 2secs
> as you still need
> to fetch that object after passing the redirector.
> That is awfully long.
>
> > I cannot avoid this though. What is the average
> time
> > for a redirector?
>
> Not sure about that as squidguards processing time
> is not measurable on
> my machine: 0ms
> I would say that less than 10ms is very good, less
> than 100ms acceptable
> - just a wild guess.
>
> > What is the maximum queue length for the
> > redirectors ? Is this a configurable parameter?
>
> No idea.
>
> > I have
> > seen FATAL errors being thrown for 22 on 10
> > redirectors. Isn't this a very small number?
>
> It depends on the number of requests, not the
> clients. Ebay is a good
> example for a site which use a lot of small objects,
> thus causing
> clients to issue a lot of small requests in a short
> amount of time.
>
> > Plz tell me if I can change this parameter some
> > place and correct the problem.
> > Regards and TIA,
> > Deepa
>
> No idea either, but I think that would be turning
> the wrong knob anyway.
> IMHO you need to improve that processing time. Can
> you implement some
> sort of caching?
>
> I dont know how perfect your remote content filter
> is and how crtitical
> bypassed request are, but this may be another
> approach: Enable bypass
> and increase the no of redirectors to find the
> minimum percentage of
> bypassed requests.
>
> Regards, Hendrik Voigtl�nder
________________________________________________________________________
Yahoo! India Matrimony: Find your partner online. http://yahoo.shaadi.com/india-matrimony/
Received on Tue Jun 01 2004 - 00:11:34 MDT
This archive was generated by hypermail pre-2.1.9 : Thu Jul 01 2004 - 12:00:01 MDT