Re: [squid-users] Web server acceleration with squid and clustering with high availability

sön 2007-01-07 klockan 15:58 +0200 skrev Forse:

> o Setup squid 1 (one that is running on same machine as web server) so
> it uses squid2 (one on VPS) as sibling and forwards all media requests
> to it.
> o Squid2 in turn will fetch media content when asked from squid1 and
> then serve it from own cache
> o Squid1 monitors squid2 and if it goes down will serve media requests
> itself until squid2 comes back online.

This is a parent relation with the default option, and with the origin
server as secondary parent without the default option.

sibling relations only borrow already cached material. You can not ask a
sibling to do stuff for you, only give you copies of what it already
has. This is the main difference between a sibling and parent relation.

> In future I want to purchase more VPS servers and create a little
> cluster where squid1 acts as main node and uses round robin to kind of
> load balance server load.

Ok.

> o My web server serves more than one website (configured to run in
> virtual host mode).

Ok

> o Both servers (web server and VPS) are running Debian unstable and
> identical versions of squid (2.6.5).

Ok.

> o I know there will probably be added latency due to squid forwarding
> requests, but main reason for me to do this is to serve media files
> from 100Mbit pipe.

Ok.

> o My squid server currently uses 2 coss cache_dir and one diskd cache_dir.

You should be using aufs.. diskd is broken..

> o VPS has 512MB ram while web server has about 4Gig. Both servers run
> on powerful hardware (dual Xeons with SCSI disks).

Ok.

Regards
Henrik