On Mon, 9 May 2011 13:48:38 +0100, John Sayce wrote:
I have two squid proxy servers. I use a PAC script to assign the
proxy servers with one being a primary and one being a failover.
This
works great but I would like to achieve a similar configuration with
the access to the WAN links from proxy servers. I have two Wan DSL
Links and two dsl routers. I'm open to changing this configuration
but I'd like to avoid a dual wan router as this would mean no
redundancy if the router fails.
Wire-level operations are *way* beyond Squid. May as well ask atoms
whether they are all spinning left or right on the third row of your
nose. The OS can do absolutely *anything* to the packets underneath
Squid. Including sending each *individual* packet out a different NIC
with a NAT'ed address Squid knows nothing about.
If you have remote upstream parent Server at the other end of each of
those WAN links then configuring them as cache_peer already does this
peer monitoring and failover at the HTTP level. Nothing special needed
beyond ICP/HTCP and ICMP. That is as close as Squid gets to link control
and if the OS decided to route to both peers via Vancouver there is
nothing Squid can do about it (my pet peeve today).
At current I have no requirement for load balancing although in
future I may assign bandwidth sensitive applications to the failover.
When you do get to this need look at QoS first. Squid can mark
connection scope to label various flows for QoS handling at the OS
level.
I could probably write a script to check the wan links and the
routers which could then change the network settings and restart if
required but this seems a rather inelegant solution. Is there a way
of doing this with squid or has anyone got any better ideas?
This is precisely how you will have to do up/down monitoring. The OS
network manager provides up/down hooks for each interface, add a script
there to reconfigure Squid when the NIC state changes.
Amos
