On 2010-05-27, Amos Jeffries <squid3@xxxxxxxxxxxxx> wrote: > > To use both you probably want to advertise both proxy IPs in the DNS > results of a name you call the cluster. The client browser can do its > own failover with several IPs when it knows. I wouldn´t trust clients to handle the failover. Would be better to have two "floating" IP-addresses that each is sticking to a different node in normal operation, but that can fail over to the other node in failure or when you want to take one node out of production. Then you do round-robin-dns over these two ip-addresses, and have a HA-LB cluster. I used to use Heartbeat for such configurations also, but have recently converted to using ucarp instead. It´s *much* easier to use, and understand than Heartbeat. > The squid can be setup as cache_peer siblings to share objects. -jf
