[CentOS] 10 Year old IT Infrastructure

Sat Oct 10 17:23:48 UTC 2009
Timo Schoeler <timo.schoeler at riscworks.net>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

thus Brian Mathis spake:
> On Fri, Oct 9, 2009 at 8:01 PM, Shawn Everett <shawn at tandac.com> wrote:
>> Thanks to everyone for their comments so far.
>>
>> The "server" in question is a basic 2 node cluster connected to an MSA500.
>>
>> It runs a variety of applications including Oracle, Apache, Samba, and a
>> proprietary app built by another vendor.
>>
>> The hardware is monitored, maintained and backed up regularly.
>>
>> The setup is mission critical to my client.  They spent a lot of time and
>> money to make sure it wouldn't go down.
>>
>> The list's point is well taken that old *nix installs are very reliable
>> long term.  I've had similar experiences.  Given this particular client's
>> need for a reliable, stable, redundant system, I was contemplating
>> alternatives or future upgrades rather than letting things age.
>>
>> Shawn
> 
> Another reason might be to avoid memory/data corruption.  Search
> Google news for a recent report from Google about how they found more
> memory errors than conventional wisdom has held to be expected.

For the archives, it can be found here:

http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf (about 300k)

http://blogs.zdnet.com/storage/?p=638

> However, consider this.  IT has a reputation among most businesses of
> always wanting new toys.  Many times, they have a point and it seems
> that IT is more interested in getting new things for no real reason.
> This sounds like one of those times.
> 
> The better solution would be to make sure you are prepared for when
> the hardware does fail.  Inform the client that you understand that
> they don't want to upgrade the servers, and that hardware failure is
> not a case of "if" but "when".  Lay out a plan to them describing what
> would happen when that occurs, and how you will make sure that their
> downtime is minimal.
> 
> My recommendation would be to look into VMware P2V conversion, and
> test it out on one of the servers if you can.  Schedule a time to run
> it once a month or so, and make sure any data that would change is
> also backed up.  You could back it up to a $99 1TB external USB drive,
> very cheap (just turn off the drive when you're done with the backup).
>  When the servers finally die, you can bring up a new server, pop the
> VM onto it, and they are back up and running.  You might not even need
> P2V if you can rsync the entire system off to the external drive.  The
> important thing here is to make sure you test bringing up the backup
> system before you're in an emergency.

Cheers,

Timo

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with CentOS - http://enigmail.mozdev.org/

iD8DBQFK0MMkO/2mgkVVV7kRAiWbAJ0ecEM0S2+BzIFxG8CVCu25b5Pd8gCgizZK
lpv0kbPFNDSJjnV+IiApA5U=
=dUhU
-----END PGP SIGNATURE-----