RAID failed again | The Coffeehouse

cheong

Today, the RAID 10 of our domain server failed again...... two harddisk of the same mirror failed at the same time without and alert emails ever sent.

I'm enough the those unreliable Promise RAID cards, and Adaptec cards are not good either. Can anyone recommand any RAID cards I can trust for?

Errata: My staffs told me that it should be Adaptec 2130S. So the current scoreboard should be 3 fails for Promise vs 3 fails for Adaptec...
(all within this 2 years)

MB

cheong wrote:
Today, the RAID 10 of our domain server failed again...... two harddisk of the same mirror failed at the same time without and alert emails ever sent.

I'm enough the those unreliable Promise RAID cards, and Adaptec cards are not good either. Can anyone recommand any RAID cards I can trust for?

Mylex used to be the only 3rd party RAID controller that I really had faith in (IBM and HP/Compaq raid controllers owed much to mylex) but since IBM offloaded Mylex to LSI Logic... I don't know what the score is.

figuerres
???

cheong wrote:
Today, the RAID 10 of our domain server failed again...... two harddisk of the same mirror failed at the same time without and alert emails ever sent.

I'm enough the those unreliable Promise RAID cards, and Adaptec cards are not good either. Can anyone recommand any RAID cards I can trust for?

Errata: My staffs told me that it should be Adaptec 2130S. So the current scoreboard should be 3 fails for Promise vs 3 fails for Adaptec...
(all within this 2 years)

Uh ?

which is it the drives or the controller cards ??

whats the working environment ??
Data center ?

are they on UPS power ?

temp control ?

brand and models of drives ?

I have setup arrays that run for 2-3 years with parhaps 1 or 2 drive losses but never 2 at the same time....

cheong

figuerres wrote:

Uh ?

which is it the drives or the controller cards ??

whats the working environment ??
Data center ?

are they on UPS power ?

temp control ?

brand and models of drives ?

I have setup arrays that run for 2-3 years with parhaps 1 or 2 drive losses but never 2 at the same time....

For the record:

It is 4 identical Seagate 73GB SCSI HDD(ST373207LW) forming RAID 10 on a Adaptec 2130S controller.

A system check reveiled the Drive1 failed, and the Drive0 is in fact ok.

The server was located in a server rack inside our company's office, with 24hrs air-conditioning and UPS attached.

figuerres
???

Ok so then how is this the card's fault?

a drive going down is out of the control of the card...

unless the card somehow killed a drive.

or perhaps you are just frustrated and venting a bit ?

cheong

I guess it could be certain bug on the card that fails the RAID when the faulty drive goes down... or perheps its not a bug.

If the faulty drive fails when the RAID card tries to read from it and sync it to the other drive, I can somehow understand......

figuerres
???

Hmmm... so a mirror broke and that cause the card's firmware / software to halt the array?

I guess this is part of why 5's are so common....

with a 3+ raid 5 with hot-spares the system just rebuilds and marks the bad drive....

but with stripe + mirror it's a bit more dicy.....

or do I have that backwards ....

mirror + stripe ??

if we have drives like this:

1,2,3 stripe as drive 0
4,5,6 stripe as drive 1

0,1 mirror as "BIG DISK"

if any one of drives 1,6 fail then the mirror has faulted and you are at risk.

been a while since I last mucked with the 1,0 or 0,1 raid modes...

cheong

Status update: After 6 hours of rescue, I successfully recovered the system.

It's time to fix those EDB errors and rebuild the RAID (I only managed to dump the partitions to another set of harddisks), but that's the work of my fellow staffs.

It's 00:32am here and I'm going home now... bye. Wink

Steven J Ackerman
StevenJ

I've been running a SCSI RAID 10 on an Adaptec PCI card under Win2K for about 5 years. I've lost two different Seagate SCSI drives, but the hot spare always worked, and I've never lost any data.

Of course, the e-mail notification never worked... I had to see the light on the array for one, and the other via a pop-up message when I logged onto the server remotely...