[squid-users] Fail-over of cache_peer parent...

I work in a small company with even smaller sales offices all over the
country with very limited network connectivity. 56k legacy leased lines in
some cases.

We have a hub and spoke network connecting all our offices.

In order to maximize the responsiveness of our intranet in the field
offices, I've implemented squid in a caching hierarchy distribution.

The problem is when the local hub goes down, I loose the cache in all the
offices relying on that hub. I'd like to set them up using the 'cache_peer'
option to fail-over to the next hub and then the home office as a last
resort, but I can't seem to figure out exactly how to do this from the FAQ's
and the configuration guide.

This is a representation of the hierarchy:

192.168.1.10 (HQ)
 10.0.1.10 (Hub1)
  - 10.0.1.11 (Field1)
  - 10.0.1.15 (Field2)
 10.0.1.20 (Hub2)
  - 10.0.1.21 (Field3)
  - 10.0.1.25 (Field4)
 10.0.2.10 (Hub3)
  - 10.0.2.11 (Field5)
  - 10.0.2.15 (Field6)
 10.0.2.20 (Hub4)
  - 10.0.2.21 (Field7)
  - 10.0.2.25 (Field8)
 10.0.2.30 (Hub5)
  - 10.0.2.31 (Field9)
  - 10.0.2.35 (Field10)

Each field office has the nearest hub as their httpd_accel_host

httpd_accel_host 10.0.1.10
httpd_accel_port 80
httpd_accel_single_host on
httpd_accel_with_proxy off
httpd_accel_uses_host_header off

Each hub has the home office apache server as it's httpd_accel_host

httpd_accel_host 192.168.1.10
httpd_accel_port 80
httpd_accel_single_host on
httpd_accel_with_proxy off
httpd_accel_uses_host_header off

Here's what I'm thinking:

cache_peer 10.0.1.10 parent 80 3130 weight=1
cache_peer 10.0.1.20 parent 80 3130 weight=2
cache_peer 10.0.2.10 parent 80 0 weight=3
cache_peer 10.0.2.20 parent 80 0 weight=3
cache_peer 10.0.2.30 parent 80 0 weight=3
cache_peer 192.168.1.10 parent 80 0 weight=4 default
prefer_direct off

Any guidance would be greatly appreciated.

Thanks,
Mike
Received on Sat Feb 14 2004 - 20:37:04 MST