The two are not really related. The redirector children value is
dependent on the speed of your redirector process, and the request load
you have. Thread count is only based on request load (and the CPU/disks
that you have).
Since you're seeing such redirector errors, you need more redirector
processes or a faster machine or a faster redirector. (Probably more
redirectors will fix it.)
The aio_queue_request message is thread related, however.
Steve Snyder wrote:
> What is the appropriate balance of redirectors used with Squid's threads?
>
> With Squid v2.3S4 I used 6 redirector children and it seemed to work well.
> When I updated to Squid 2.4S1 (+ posted patches) I upped that count to 8
> redirectors, now in use with the 16 threads handling the aufs async I/O.
>
> Today I noticed these warnings in my log file:
>
> WARNING: All redirector processes are busy.
> WARNING: 8 pending requests queued
> Consider increasing the number of redirector processes in your config file.
> aio_queue_request: WARNING - Queue congestion
>
> So... what is the appropriate balance of redirector children to aufs
> threads? Do I need a redirector for each thread (16 by default) created by
> Squid v2.4? I've got Squid running on RedHat v7.1, BTW.
>
> Thank you.
--
Joe Cooper <joe@swelltech.com>
Affordable Web Caching Proxy Appliances
http://www.swelltech.com
Received on Wed Jul 18 2001 - 16:02:15 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:01:10 MST