Ow Mun Heng wrote:
In theory.. I think I need to experiment on this then.
You do not want to do this...I saw someone say you wanted to be on the
same subnet; I wouldn't even do it that way if it were me. I would run
serial cable direct or I would run crossover cable directly to
interfaces. The last thing you want is heartbeat flapping due to
latency. Usually heartbeat works well, but I have seen it bring up all
services/interfaces on all the servers which were failover for one
another. It is possible, and as you can imagine, it's not a case you
want a production environment to be in.
Seeing that you're a sys-admin for managed hosting, do you have any tips
on load-balance for multiple thousands of squid users?
We use LVS (Linux Virtual Server) as our loadbalancer at the moment.
www.ultramonkey.org and www.linuxvirtualserver.org
What's the likelihood that squid will fail and then users are left w/o
any I-net connections?
It's going to fail - it is just a matter of when.
Andrew
Andrew
eg:
---[center1]--interstate--[center2]---
center1 and center2 will have independent squid installations. (but each
will use the other as sibling)
most likely using transparent proxy/wccp/ bridging (which one is better)
and I would like users connected to center1's squid be fail-overed to
center2's squid in case center1 goes down. (otherwise, being a
tranparent proxy, if center1(or 2)'s squid goes down, (for some reason)
I-net access for that center will fail.
AFAIK, Hearbeat uses serial cables for monitoring, (I may be wrong, the
last time I read the documentation).
I would appreciate any help in this.
--
Ow Mun Heng
Gentoo/Linux on D600 1.4Ghz
Neuromancer 18:00:01 up 8:28, 6 users, 1.64, 0.76, 0.54
--
Ow Mun Heng
Gentoo/Linux on D600 1.4Ghz
Neuromancer 11:07:06 up 1:45, 6 users, 0.76, 0.54, 0.58
