Re: [squid-users] Too many queued redirector requests

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 Voigtländer
Received on Mon May 31 2004 - 08:51:53 MDT