Re: [squid-users] Multisite ICP peering

From: Chris Toft <>
Date: Wed, 3 Nov 2010 19:53:48 +1100

Thanks for the reply, I actually fixed it. Removed the multicast-responder option and just left multicast-sibling.

Man this thing flies on 5 boxes with 64gb memory and 10x 50gb solid state drives for the cache :-)

I will post working config tomorrow for anyone interested.

Chris Toft
Mob: 0459 029 454

On 03/11/2010, at 19:51, Josip Almasi <> wrote:

> Chris Toft wrote:
>> Hi,
>> I am having an issue with getting multisite ICP peering to work correctly.
>> Let me lay out the scenario.
>> Two sites, one primary and one DR.
>> 3 squid caches on the primary site, 2 on the secondary site in front of a cluster of 4 back end web servers on each site.
>> Primary Site
>> I have set them up so when requests hit the squid caches (behind a loadbalancer) on each site they will check their squid peers for the image before passing to the back end. This is exactly what we wanted.
>> Secondary Site
>> Again setup the same as the primary site but in the configs all 5 squid caches are listed as multicast peers, but when the secondary site is hit it only checks the caches on that site and then goes to the origin webservers.
>> The idea being the primary site will get hit 90% of the time and therefore have a "hot" cache, with 10% of traffic hitting the secondary site to give a "warm" cache. This being the case we wanted the secondary site squid caches to use the multicast peers in the primary site before hitting the backend webservers.
>> I have tried various options of multicast, sibling, parent for the type of server.
>> I have tried using weighting to make the primary site caches a higher priority than the backend web servers
>> There are no connectivity issues because if I comment out the secondary site webservers in the configs then it will try the primary site squid caches and then fall back to the primary site webservers.
>> How can I setup the config so that when a request comes in to the secondary site it will check the primary site cache for an object before falling back to the webservers???
> You got to have class D address to get UDP multicast working, or mbone,
> or VPN/private addresses.
> Alternative is cache digest, just shorted the time period. And then I'd
> recommend using 'proxy' instead of 'sibling' on sec site.
> Regards...

The information contained in this e-mail message and any accompanying files is or may be confidential. If you are not the intended recipient, any use, dissemination, reliance, forwarding, printing or copying of this e-mail or any attached files is unauthorised. This e-mail is subject to copyright. No part of it should be reproduced, adapted or communicated without the written consent of the copyright owner. If you have received this e-mail in error please advise the sender immediately by return e-mail or telephone and delete all copies. Fairfax does not guarantee the accuracy or completeness of any information contained in this e-mail or attached files. Internet communications are not secure, therefore Fairfax does not accept legal responsibility for the contents of this message or attached files.
Received on Wed Nov 03 2010 - 08:53:49 MDT

This archive was generated by hypermail 2.2.0 : Tue Nov 09 2010 - 12:00:02 MST