I have a T420 with Perc H310, 4 drive bays, configured as two RAID 1 disk groups, with 12 virtual disks configured on them
I wanted to replace one disk group with larger 4TB disks to provide more space for additional virtual disks.
The manual describes the Replace Member operation that looks as though it would accomplish this task, but that's not feasible because there is no spare slot to contain the new disk.
So what I did was break the mirror, remove one disk, put the new larger disk in the same slot, configured it as a global hot spare, and then rebuilt the mirror. That was fine. Now I repeated that procedure, this time replacing the other original disk with a larger one and again rebuilding.
Thus I now have the same disk group, all nicely mirrored with all the same data, but with larger disks. However the disk group still only shows as having the original capacity, and doesn't allow creation of new virtual disks. In fact the disk group doesn't permit any operations at all, so the Perc is obviously not happy with what I've done, even though it's actually working just fine.
So what should I have done? I can't see there were any other options available to me, other than starting afresh with an empty disk group and restoring the whole thing off backups, which I thought should be unnecessary- surely the point of RAID hardware is to make this sort of thing easy?
In fact I did exactly the same operation with a Synology DiskStation recently: remove one disk, put in a larger one, rebuild, remove the other disk, put in a larger one, rebuild, job done - simplicity itself. Why should Dell be any more complicated?
If the answer is 'sorry, restore from backup is the proper way', then what about this for a plan?:
1. break the mirror;
2. remove one disk;
3. insert larger disk, and configure it as RAID 0 (ie a new disk group);
4.create identical virtual disks on new disk group;
5. copy original virtual disks to new ones (in operating system, which is Hyper-V Server 2019);
6. delete original disk group and remove disk;
7. insert new large disk, make RAID-capable;
8. convert large disk group to RAID 1 using spare disk.
I don't see why this shouldn't work.
I should perhaps point out that simply restoring from backups is anything but simple, because most of the virtual disks contain VHDs for virtual machines, backed up on iSCSI targets on the DiskStation, with the backups orchestrated by Windows Server Essentials. In other words, the backup strategy operates at a level way above the hardware level: I'm sure it could be done, but it would be a nightmare trying to do it.
If the H310 supports that feature then it is called Online Capacity Expansion. It would be performed by using something like OpenManage Server Administrator. In OMSA there will be a reconfigure option on the virtual disk. If OCE is supported on the storage controller and the virtual disk meets requirements for OCE then there will be an option to expand the virtual disk capacity.
I don't think options like retag are possible on the H310. I think it forces an initialization when a VD is created. You can find more information in the storage controller manual and the OMSA storage manual.
Dell EMC, Enterprise Engineer
Thanks for taking the time to reply, Daniel.
Yes, the H310 does support Online Capacity Expansion, but it appears to be only via the Replace Member function, and that is not possible if there are no spare drive slots. The manual contains this statement:
'NOTE: If the controller already contains the maximum number of virtual disks, you cannot perform a RAID level migration or capacity expansion on any virtual disk'.
That is a very peculiar statement. Does it really mean 'the maximum number of virtual disks', or should that be 'the maximum number of physical disks'? The latter seems to make more sense.
In any case, I don't want to expand any of the virtual disks, I just want the spare capacity in the larger drives to be available for creating new virtual disks.
The H310 doesn't appear to support retag.
It seems to me that the best way forward is to try the procedure I suggested in my initial post, unless anyone can think of a reason that won't work.