Re: Transperant HTTP Redundancy.

Goncalo Valverde wrote:

> failure.. sometimes when this happens the machine still
> replies to pings, squid is up and running, but obviously doesnt
> replies to requests.. so you could grab the ip address because the
> other machine is still using it... or is there any way around this
> problem?

There is always ways around a problem...

Problem:
How to build a hign-availablilty server using a stand-by server. The
standby machine should take over when the main machine stops processing
requests, whatever reason that is.

Solution: Use some kind of montitoring to tell when the machines should
be active on the network. The monitoring needs to include accurate tests
of the montitored service and automatically reboot the machine on
un-recoverable errors.

Special cases:
* Machine is alive, but the process has died/halted.
- Restart the service (Squid). If this takes to long time or fails then
shut down the IP alias and signal the stand-by machine.

* Machine is half-dead: responding to IP traffic, but nothing processed
or can't recover
- Something needs to reboot the failing machine, or at least disconnect
it's network connection. Use a watchdog timer to automatically reboot
the machine if it has problems.

* Machine is dead
- Can easily be detected by the stand-by machine.

And you probably should have and some way to recover if both should
become active. This can happen if there is a external failure (for
example a network failure) causing the monitoring to beleive that the
other machine is dead.

---
Henrik Nordström
Sparetime Squid Hacker