Hello Toaster Administrators,
We run a 3170 in front of another vendor's disks and recently had a panic/failover event where we didn't get any crash dump files because there was no spare disk available to save to.
Technically the 3170 has no spare disks because it runs RAID0 on all the "disks" it sees presented from the SAN (a 3Par T800) and redundancy is all managed on the T800.
NetApp has told me we need to have some spare disks available. I've provisioned two 20GB "disks" and exported them to the 3170. Seems like all I need to do now is assign them (auto assign is not on), one to each filer node.
But am I correct that the next time I carve out real (bigger) space and go to add it to some aggregate, those crashdump disks will also be available to add? This makes me cringe a little as we're in the habit of using the defaults on the "add array LUNs" wizard including any size disk and we confirm that the number of available disks is the same as whatever we just deployed for the addition. To avoid using the dump disks, we'd need to (as one idea) filter the wizard to only use certain size disks. This is really just a training issue but I can still see the day clearly when someone spaces and adds a 20GB disk to an aggregate that then can't be removed. Messy messy messy....
Is there a way to "hard" assign a disk as a spare? So that it won't be available for addition to an aggregate later?
Hope to hear from you...
Randy Rue Fred Hutchinson Cancer Research Center Seattle, WA USA
My opinion is to deploy 1 DS14 style shelf with each set of V3170 as long as you can afford to lose 2 fc ports on each controller (or go with single path ha.. ymmv). Then you can carve up the 14 disks as 7 to the A side, and 7 to the B side, create your root aggr / vol on 6 disks and leave one disk as a spare. Giving up prime and possibly very expensive 3rd party storage just for a possible core dump just seems wasteful, but necessary if you have nothing else.
It's hopefully not much money to give up for better supportability. In our environment, we're able to give up 1 shelf off older equipment for just such a purpose. It works great, and we get a good piece of mind that if/when something terrible happens, we'll have logs, and hopefully a core dump for analysis.
-Blake
On Mon, Apr 11, 2011 at 2:21 PM, Robert McDermott rmcdermo@fhcrc.org wrote:
Hello Toaster Administrators,
We run a 3170 in front of another vendor's disks and recently had a panic/failover event where we didn't get any crash dump files because there was no spare disk available to save to.
Technically the 3170 has no spare disks because it runs RAID0 on all the "disks" it sees presented from the SAN (a 3Par T800) and redundancy is all managed on the T800.
NetApp has told me we need to have some spare disks available. I've provisioned two 20GB "disks" and exported them to the 3170. Seems like all I need to do now is assign them (auto assign is not on), one to each filer node.
But am I correct that the next time I carve out real (bigger) space and go to add it to some aggregate, those crashdump disks will also be available to add? This makes me cringe a little as we're in the habit of using the defaults on the "add array LUNs" wizard including any size disk and we confirm that the number of available disks is the same as whatever we just deployed for the addition. To avoid using the dump disks, we'd need to (as one idea) filter the wizard to only use certain size disks. This is really just a training issue but I can still see the day clearly when someone spaces and adds a 20GB disk to an aggregate that then can't be removed. Messy messy messy....
Is there a way to "hard" assign a disk as a spare? So that it won't be available for addition to an aggregate later?
Hope to hear from you...
Randy Rue Fred Hutchinson Cancer Research Center Seattle, WA USA
-
Blake Golliher -
Robert McDermott