Re: High availability

On 26 Jan 2006, in the Usenet newsgroup linux.redhat, in article
<1138277982.474502.51070@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, larsk wrote:

>I am building this portal, that needs high availability and I am
>thinking about clustering the 3 linux boxes to be one so if one box
>goes down the other two will still keep the site up.

-rw-r--r-- 1 hm linux 130414 Dec 25 1998 High-Availability-HOWTO.tar.gz

-rw-rw-r-- 1 gferg ldp 26237 Jan 8 2005 Beowulf-HOWTO
-rw-rw-r-- 1 gferg ldp 17964 Sep 5 20:35 Cluster-HOWTO

You're probably going to have fun finding the High-Availability-HOWTO, as
I don't see it on sunsite (was in /pub/linux/ALPHA/linux-ha/)...

Web Results 1 - 10 of about 64 for High-Availability-HOWTO.tar.gz.
(0.88 seconds)

(OK, it's out there) but that probably provides a very good base to start
from.

If the site is so critical that you are worried that you need three
systems clustering it, you probably want to consider (at the very
least) multiple power sources, multiple network links to whereever your
client/customers are, and multiple physical locations located in different
areas in different flood plains, and so on. Assuming you made reasonable
hardware selections, the only time the individual systems should be down
is when you reboot them after replacing the kernel or a few critical
system libraries.

>does this sound good or does anyone see any holes in this?

Depends on what your threat scenario is.

Old guy
.