Q about NAS mounting things - Mailing list pgsql-general

From Joel
Subject Q about NAS mounting things
Date
Msg-id 20041117173107.C35B.REES@ddcom.co.jp
Whole thread Raw
List pgsql-general
Long and rambling question, here -- we have a customer with a database
that is outgrowing their original storage, and they want to upgrade
storage as cheaply and smoothly as possible.

Presently, they have one 120G or so RAID 5 unit, and the entire system
and data are stored on that drive.

We've talked about cloning the current drive to a larger capacity RAID
and swapping the larger one in. We've also talked about hanging a second
(external) RAID 5 unit on the machine and just copying the data over,
leaving the OS, the system, and the software components on the original
drive. These solutions are perceived by the customer as too expensive or
something.

Now we are considering NAS, mounted over NFS. (So far they have only
been able to find a Terabyte class cheap NAS running MSWindows, but I
think we're going to look harder for Linux NAS.)

I've suggested running postmaster on the NAS, but we don't seem to want
to dig into the code.

One of our group wants to copy both data and postgresql to the NAS,
running postmaster on the server from the binary saved on the NAS. I
think this sounds too fragile. Specifically, if the NAS goes off line
while postmaster is running, it will tend to go looking for code and die
the horrible, won't it?

So I have recommended data only on the NAS, modifying the startup and
shutdown scripts to check the presence of the NAS before attempting to
mount, and adding a watchdog to check every second whether the NAS is
still mounted. I'm also expecting we'll be able to set postgreql to fire
off an e-mail if it suddenly can't find the files because the NFS
mounted NAS has disappeared for some reason.

Opinions, anyone?

--
Joel Rees   <rees@ddcom.co.jp>
digitcom, inc.   株式会社デジコム
Kobe, Japan   +81-78-672-8800
** <http://www.ddcom.co.jp> **


pgsql-general by date:

Previous
From: matthias@cmklein.de
Date:
Subject: Re: Problems importing Unicode
Next
From: Richard Huxton
Date:
Subject: Re: Problems importing Unicode