We just installed a FAS8020. It is dual controller with two shelves of 900 GB SAS drives and one shelf of 4 TB drives. The plan is to use the 900 GB SAS drives for production systems and the slower 4 TB drives for backup and dev and test systems.
My plan was to put the two 900 GB shelves on controller 1 and the 4 TB shelf on controller 2. I'm torn about having a dedicated root aggregate - I would like to have as much space as possible and it seems like more and more folks are moving away from that so I would like a recommendation. I will have 1 spare of each drive.
Initially the system came up with a dedicated root aggregate on controller 1 with three of the 900 GB drives and a dedicated root aggregate on controller 2 with three of the 4 TB drives.
Should I forget about dedicated root aggregates and just have a single aggregate for each controller?
If not, should I recreate the root aggregate on controller 2 with three of the 900 GB drives so I'm not using 12 TB for a root aggregate? If so, how do I get a root volume on a new aggregate?
As I see it have the following options:
1) A dedicated root aggregate on controller 1 and another aggregate of 44 drives with 1 spare - all 900 GB drives. A dedicated root aggregate on controller 2 and another aggregate of 20 drives with 1 spare - all 4 TB drives.
2) A dedicated root aggregate on controller 1 and another aggregate of 41 drives with 1 spare - all 900 GB drives. A dedicated root aggregate on controller 2 with three 900 GB drives and another aggregate of 23 drives with 1 spare of 4 TB drives.
3) No dedicated root aggregates. Controller 1 is one aggregate of 47 drives with 1 spare - all 900 GB drives. Controller 2 is one aggregate of 23 drives with 1 spare all 4 TB drives.
It does seem that people are moving away from dedicated root aggregates. This is mainly due to having large disks and not desiring to "waste" the space (I know you can "use" the extra space in the root aggregate, but that's basically the same as not having a dedicated root aggregate, with lousier performance). We moved away from dedicated root aggregates years ago and we've not had a single negative impact. Personally, I would choose your option 3, except I would probably leave 2 spares on both controllers (especially controller 1 with 2 shelves). From my experience, even though we have premium support, I sometimes don't get replacement disks installed in a timely manner. A remote filer pair might be in a building closed during the weekend and no one's there to replace the disk or allow an engineer in to replace the disk, etc.
We're in a 24x7 manned facility so as long as I know a drive is down I can get it replaced. We even have talked about having an extra drive of each type on the shelf. Having two spares of 1 TB drives would be doable but I'm a little tight on the 4 TB space. But if I go without a dedicated root I pick up two drives anyway.
If I went this route would you just add the disks to the existing root aggregates? Do I need to do anything about RAID group size?
Milage varies, personally on small systems I don't use a dedicated root aggregate, on bigger ones I do becasue its a nice place to tuck away my vol0's for vfilers, any huge logs I need to hang onto for a bit (like packet captures) and becasue our backup guys seem to have trouble with the concept of restroing to anything otehr than "/"
In your case I'd probably create a 3 disk root aggregate out of you 900 gig drives (use "aggr options <aggrname> root" to make the aggregate root at next reboot)
if you move aggregates don't forget to also move vol0 (vol options "volname" root for that one at next reboot too - though most people I know just rename it to vol0 so we don't confuse the support guys)
But if using 3 900gig drives dosn't float your boat...just go without a dedicated root aggregate.
Yes, if you go this route just add disks to the existing root aggregates. For raidgroup size I would recommend going as large as you can, but remaining balanced. For example, for your 24 SATA drives, assuming you're going with raid-dp and keeping in mind you only want one spare and max capacity, I would go with rg size 12 (with rg0 having 12 disks and rg1 with 11 disks). For your 48 SAS drives on controller 1, I would go with rg size 24 (with rg0 having 24 disks and rg1 with 23). YMMV, these are of course subject to the size limits of your hardware model and version of Ontap.
Keep in mind that you cannot decrease the rg size, so be careful. I would also recommend checking the layout of the disks being added to the rg's before you commit to it (use the CLI for full control, or if using System Manager, go into the raid details to see the Raid Group Layout).
Thanks for your responses and advice. I was going to go without the root aggregates but my colleague argued we should convert the 4 TB root aggregate disks to 900 GB root aggregate disks because we can afford the loss of 4 disks space wise and the root aggregate is a "best practice". That's a problem in our organization where we get beat up if we deviate from best practice and something goes wrong - even if it is unrelated. In the end I did 40 SAS 900 GB drives in an aggregate with a RAID group size of 20. The 23 SATA drives are in a RAID group size of 12.
It's cool. You do what's best for your business requirements. The thing with "best practices" is that they become dated real quick and usually don't get updated. What may have been best practice when 300GB disks were the norm may no longer be best practice today. Have fun with your filers