On Wed, Jan 06, 2010 at 02:11:10PM -0600, Les Mikesell wrote:
> On 1/6/2010 1:17 PM, Stephen Harris wrote:
> > All hardware sucks, all software sucks.
> >
> > Your machine _will_ go wrong and you _will_ need remote console access and
> > remote power ability. Especially if you have thousands of these things.
>
> Of course things break - including the extra stuff you might add for out
> of band access, but (a) many/most of the things that break can't be
> fixed remotely so you are going to end needing to swap things out
That wasn't my experience when I was an SA.
> anyway, and (b) if you have thousands you should have enough redundancy
> to survive until you get around to swapping the broken thing with
> something that works. An IP KVM might save a trip or hands-on support
Nope; these machines are all used for different applications, by
different sub-lines of business, and are not sitting there idle.
If you have 1000 boxes all doing the same thing (hi, Google!) then
sure; if you have 1000 boxes all doing _different_ things then no.
> call once in a while but the odds aren't that great.
When I used to be on call (fortunately not for 9 years, now) and when
I got paged for a problem, I'd estimate 1 in 20 of the calls required
a direct root login, and that was limitted to the console. Some of the
problems could stop a normal user logging in (eg automount failure; NIS
binding issue; network routing issue). Some of those required a physical
presence, but the majority of them could be fixed from the remote console.
Heck, even for scheduled changes (bring server up single user, networking
not active) remote consoles made the work possible.
Remote consoles (whether it be a IP KVM, or an old style Sun Sparc serial
console) aren't luxuries for a large environment; they're essential.
--
rgds
Stephen
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
