Scott,

Excellent question.  I assumed, possibly incorrectly, that cost would be major criteria given the tight budgets at our institutions of higher learning.  Clustering and mirroring with Vfilers are excellent solutions, but they tend to press on limited budgets. 

 

A pure hardware solution, while not elegant, has minimal cost, provides very short MTTR and minimal operating complexity.

 

With the limited number of shelves on each system a clustered environment may require more shelves and associated disks to preserve a boot OS environment on both machines.

Hunter

 

Hunter M. Wylie

21193 French Prairie Rd

St. Paul, Oregon 97137

Bus: 503-633-8900

FAX: 503-633-8901

Cell: 503-880-1947

-----Original Message-----
From: Waters, G Scott DSTI [mailto:george.scott.waters@us.army.mil]
Sent:
Friday, April 25, 2003 12:06 PM
To: 'Hunter Wylie'; 'Michael Homa'; 'toasters@mathworks.com'
Subject: RE: Duplicating filers

 

Why not just look at clustering the two filers??  ... or am I missing something?

- Scott W.

-----Original Message-----
From: Hunter Wylie [mailto:hwylie@stpaultel.com]
Sent: Thursday, April 24, 2003 2:36 PM
To: 'Michael Homa'; toasters@mathworks.com
Subject: RE: Duplicating filers

 

Michael,
In theory it might work on a "perfect" day "but" the number and type of simple failure scenarios would make the perfect day highly improbable. I suggest you take a simpler approach that many of my very large customers use to achieve  very high availability, very short mean time to repair, simplicity and low cost.

Buy a used F740 (head) configured like the biggest one you have, a spare 36 and 18 GB disk and a FC9 disk shelf.  When something breaks and you have an idea it's "head associated - down the dying F740 and power it off, unhook LAN connections, FC disk cable, etc. and attach them to the spare F740 racked between the original two heads. Cable lengths can get you if you are not careful.  Watch your power load on the rack/circuit.

If it is shelf associated, shut down the system, power it down, remove the old shelf cable and attach the new one, move all the drives and power up.

In both cases the system will comeback on line with all the old identities re-established, no DNS hacks. 

This technique is used by one of the largest on-line retailers in the world. They love it.

No software licenses are needed on the spare system as long as it is "NEVER" used in parallel production with the two other systems. In this

mode it is strictly a replacement spare part.   If you want to use it in
any other mode you must acquire software licenses with it.

The nice thing about this solution is it is simple to test.

Send me an email if you want to figure out the exact parts you will need.

Hunter M. Wylie
21193 French Prairie Rd
St. Paul, Oregon 97137
Bus: 503-633-8900
FAX: 503-633-8901
Cell: 503-880-1947

 

http://www.itmls.com/

 

-----Original Message-----
From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]
On Behalf Of Michael Homa
Sent: Thursday, April 24, 2003 10:24 AM
To: toasters@mathworks.com
Subject: Duplicating filers

Hi.

My name is Michael Homa and I'm a member of the Systems Group at the University of Illinois at Chicago.

Some background
---------------

We have two F740 filers:  netapp1 and netapp2.
  o Both of them are at ONTAP 6.2.1.
  o Each has a different admin host.
  o Both have two FC9 data shelves. On netapp1, one shelf contains 18GB
    drives; the other shelf, 36GB drives. On netapp2, both shelves
    have 36GB drives.
  o Netapp1 has one volume, vol0, and two raid groups, rg0 and rg1.
  o Netapp2 has two volumes, volboot and vol0, and each volume has a
    single raid group, rg0.

Netapp1 is the "center of our universe" and is NFS-mounted to almost all of our production servers. Netapp2 currently functions solely as the recipient of the snapmirroring of netapp1.

I've been asked to undertake a project in which I make netapp2 be a duplicate of netapp1. In the event that netapp1 becomes unavailable (maybe a catastrophic hardware failure), we want to re-cable netapp2 to the network, monkey with DNS, and reboot the box. In effect, for all intents and purposes, make it netapp1.

It's a good idea in theory but maybe in practice it won't work out. Nope. Now that I think about it, it doesn't even sound like a good idea in theory but my boss has asked me to try it so what the heck.

Has anyone attempted this? I'm looking for suggestions, ideas, "gotchas." I don't need to reinvent the wheel; I'm more than willing to learn (borrow, steal) from others.

Thanks.

Michael Homa
Academic Computing and Communication Center
University of Illinois at Chicago
email:  mhoma@uic.edu