Ok thanks for clearing it up.  Load balancing I can do on my own
but HA is what I need, apparantly.  I still have my terminologies
confused. 

DNS TTL tricks wont still do so for what i want.  Right now
I guess we'll just settle for backing up just the MX traffic and
just have to work out driving over to the collocated server to
fix things (30 minutes to 1 hour drive).

All the other ideas are cool--thanks for the info.

jondz


On Sun, 2003-07-06 at 20:13, Rafael 'Dido' Sevilla wrote:
> On Sun, Jul 06, 2003 at 06:36:06AM -0400, Hagibis Fan wrote:
> > No, I wont go with round-robinning
> > DNS A entries--all traffic should go to
> > the main server at all times but will
> > fallback to our other servers when thats 
> > down.  Playing around with TTL or any
> 
> What you're describing hear is *not* load balancing, but high
> availability service.  Load balancing will do something totally
> different from what you describe, it will balance the load among your
> entire pool of servers, not concentrate traffic to a single server and
> then fallback to the other servers if that goes down.
> 
> > other DNS trick wont do either.  Either
> > its intantaneous or I wont do it.
> > 
> 
> Whether this is possible or feasible depends on the lengths you're
> willing to go and who will have access to your services.  If your
> primary and backup servers live on different networks and you want your
> service to be accessible to the Internet at large, well, I'm afraid to
> say that you cannot achieve instantaneous failover.  The only way to do
> it is round robin DNS and/or TTL tricks.  If you're concerned about
> accessibility of your website to assorted locations, and have the money
> to spare, buy service from Akamai.  The largest websites like Yahoo and
> CNN use their system.
> 
> If you have a closed set of clients for your web service, you can set up
> a VPN for your primary and backup servers and your clients and then run
> Heartbeat (http://www.linux-ha.org/) between your primary and backup
> servers.  Assign a cluster IP address that passes between your primary
> and backup servers.  Although the docs for heartbeat say you should be
> running the heartbeat over a dedicated ethernet or serial line, there is
> no reason why you can't run it over your normal network interface as
> well.  That will provide near instantaneous failover.
> 
> > I've tried to read up BGP but I cannot
> > understand it--that technology will be 
> > beyond me.
> 
> BGP won't help you here.  You aren't trying to do IP load balancing at
> all, and BGP is intended to load balance arbitrary traffic from two
> different upstream sources to a single destination.  It's *totally*
> unrelated to server load balancing, much less high availability.
-- 
Hagibis Fan <[EMAIL PROTECTED]>

--
Philippine Linux Users' Group (PLUG) Mailing List
[EMAIL PROTECTED] (#PLUG @ irc.free.net.ph)
Official Website: http://plug.linux.org.ph
Searchable Archives: http://marc.free.net.ph
.
To leave, go to http://lists.q-linux.com/mailman/listinfo/plug
.
Are you a Linux newbie? To join the newbie list, go to
http://lists.q-linux.com/mailman/listinfo/ph-linux-newbie

Reply via email to