diskless or something like that (read...please...)

[999364]

Hi,

I need some help...

my bosses ask me if exists something like this:

They decided to have 2 machine (1,2),
1st - there are 3 application running on separate RAID5 (I mean 3 RAIDS).
2nd - a spare machine that is connected only if 1 of the previous raid fault (I know that RAID5 is fault tollerance, but let imagine that the controller faults, so HDs are fault and 1 of the application is down).

They would like to connect the spare part not manually (like hot-swap) but using a switck (may be by LAN).

They do not want cluster (they says is too expensive).

Do yuo have any suggestions, or some docs, links where to find something useful....

Do not hesitate to ask more to understand my issue....!!!

thank a lot

 

[iolair]

Well, what you are asking about is basically what a failover cluster does. If you're using Windows 2003, it comes with clustering standard, so I can't see what the expense would be. If you want to actually just make the disks available, have you considered a SAN or even NAS?

Iolair MacWalter

Iolair MacWalter
Director of IT

 

[999364]

I spoke about SAN and NAS, win srv 2003 cluster but they want to keep this hardware architecture....

Do you any suggestion....

thank

 

[dkediger]

Unless they are the Dept of Defense (or funded by them), there's pretty much always going to be single point of failure somewhere in the system. What if the switch dies for example.

Do they realize that 3 RAID5 arrays is at a minimum 9 disks per machine? What's the availability/cost of a server with 9 disk slots?

With 3 apps, I would rather utilize 1 machine per app, ideally with a SAN. With a SAN out of the question, then depending on the value they put on uptime, I would RAID 0+1 or 0+5 each machine.

 

[999364]

Can you explain better what you mean...
thank

 

[WhoKilledKenny]

Dude,
Welcome to the world of bosses we call dumb asses!!! I feel your pain, I too work for a dumb Ass. They want uptime of 99.9999% on a 90% budget, they will never get it. If they want to use the existing hardware without spending the extra money to do a proper cluster solution, then this is your only solution. Use the second box as a cold spare. If the prduction server goes south there will be downtime to bring the second box online, configure IP, and change the network name. It all depends on how you set up your disater recovery plan for that system.

Just a clarification on iolair's post, Windows 2003 Standard allows NLB clustering (Network Load Balancing). To do a Microsoft cluster with fail-over resources, disk, you know... the whole shaaabang - requires Windows 2003 Enterprise Edition.

Good luck homie...