[rescue] Enterprise 3000/3500 and StorEdge A5000 rescue

[Mr Ian Primus]

--- On Fri, 9/24/10, Brian Deloria <bdeloria at gmail.com> wrote:

> I only used the 9GB standard ones in years past. 
> Formatting went fine under
> SuSe years ago, about the only hitch was I think on one of
> the A5200's which
> has 22 disks that when you plugged the second controller up
> it ran out of
> /dev devices about half way through.

I'm running three of the 14 disk arrays. When all connected, the system goes well past the end of /dev/sd[x], and goes up to /dev/sdaa, /dev/sdab, and so on. I'm running Debian. Things work well, but something is seriously screwed with the way the kernel detects partitions at startup...

If you partition the disks with Solaris-style disklabels, you have the ever-present whole-disk partition as the third partition. Then, you create the single partition of Linux RAID AutoDetect as the first partition, and build your RAID using /dev/sdb1, /dev/sdc1, etc. Unfortunately, since you have used solaris disklabels, /dev/sdb3p1 is technically another way to refer to /dev/sdb1, so when the kernel goes to bring up the raid devices, it keeps finding all these partitions with the same superblock! Shock and horror! So, you either have to plug in the fibre after the system has already booted and manually reassemble the array, or just redo all the disks without the whole-disk partition and be done with it.

-Ian