Re: [squid-users] Response error from squid when runnng bench mark.

From: Mark Elsen <mark.elsen@dont-contact.us>
Date: Sat, 15 Apr 2006 08:44:58 +0200

> Hello,
>
> I am running the polygrpah test on squid. My setup
> with NO-PROXY runs just fine.
>
> But when I run with proxy I am getting the following
> error, though the test seems to complete with out
> stoping in the middle.
> ======================================================
> 002.46| i-warm 376 3.00 1989 0.00 0 188
> 002.54| i-warm 387 2.20 2269 0.00 0 194
> 002.63| i-warm 407 4.00 2023 0.00 0 195
> 002.71| i-warm 422 3.00 2644 0.00 0 194
> 002.79| i-warm 430 1.60 2299 0.00 0 193
> 002.88| i-warm 445 3.00 1890 0.00 0 194
> 002.96| i-warm 467 4.40 2244 0.00 0 198
> 003.04| i-warm 478 2.20 3224 0.00 0 196
> 003.05| fyi: servers visible: 513 reached: 210
> (40.94%) attempts: 760
> 003.13| i-warm 493 3.00 2582 0.00 0 197
> 003.21| i-warm 508 3.00 2437 0.00 0 196
> 003.22| Connection.cc:485: error: 1/1 (s110)
> Connection timed out
> 003.22| error: raw write after connect failed
>
>...

  The intend of a polygraph setup is that your robots and severs
should not run on the same box where SQUID resides.

The SQUID box, is the taget of the test however, since in a correct setup,
all requests to the servers from the robots goes thru SQUID.

Polygraph, poses firm load on the 'network stack' of any machine,one
of the things to look at is that the ip local port range, may need increasing
, for instance.

Any quick search on GOOGLE with those keywords, will return links to
you, as how to do that.
FD's and max open files, may also need to be increased.

Also checkout :

         # dmesg

on the nodes where the robots are trying to do their work, I once
got "Out of socket memory" ; no so ease to resolve, as it may need
further kernel changes and or changes of kernel parameters.

M.
Received on Sat Apr 15 2006 - 00:44:59 MDT

This archive was generated by hypermail pre-2.1.9 : Mon May 01 2006 - 12:00:02 MDT