Re: [squid-users] source-hash balancing...

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Tue, 2 Sep 2008 14:31:03 +1200 (NZST)

> On Mon, 1 Sep 2008 07:34:37 -0700 (PDT)
> John Doe <jdmls_at_yahoo.com> wrote:
>
>> Cool, thx.
>> Would the following work...?
>>
>> # u1 servers pool
>> cache_peer 192.168.16.101 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u1pool
>> cache_peer 192.168.16.102 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u1pool
>> cache_peer 192.168.16.103 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u1pool
>>
>> # u2 servers pool
>> cache_peer 192.168.16.201 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u2pool
>> cache_peer 192.168.16.202 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u2pool
>> cache_peer 192.168.16.203 parent 80 0 no-query originserver no-digest
>> no-netdb-exchange max-conn=256 sourcehash name=u2pool
>>
>> acl u1 url_regex ^http://u1
>> acl u2 url_regex ^http://u2
>> cache_peer_access u1pool allow u1
>> cache_peer_access u1pool deny u2
>> cache_peer_access u2pool allow u2
>> cache_peer_access u2pool deny u1
>>
>> Won't there be a problem with the redundant 'name=u?pool'

Yes. The name= option is a UID for each peer. Also regex is very very
slow. dstdomain is better for those ACL.

>
> I tried it once with squid 2.6. It did not work. But I would really
> like it if that would actually work (i.e. grouping multiple peers
> together so one doesn't need to create the same cache_peer_access-rules
> for all peers).

Good idea. They are already done loosely that way for selection methods,
but there does not appear to be anything to group peers based on a tag,
and certainly nothing to group *_access lines together yet.

Want to spec out a 'pool=' option for squid?

Problems:
 How should a group with mixed selection methods be handled?
 How should specific per-peer ACL affect peer group ACL?

Amos
Received on Tue Sep 02 2008 - 02:31:10 MDT

This archive was generated by hypermail 2.2.0 : Tue Sep 02 2008 - 12:00:02 MDT