SIZING 24 SSDs for 2 FLASH POOLS

[heightsnj]

I have 1 shelf SSD X24. planning on creating 2 flash pools for 1 agg on each node within a pair, RAID-DP

Should I do: 9 data, 2 dp, 1 spare, or 8 data, 2dp, 2 spares for each pool? can data disks be odd number?

Also, among 24 SSDs, 7 is owned by one node, 17 is owned by the other. Should I split the ownership by half and half before creating the pool or does that matter?

Thanks for sharing!

[rwelshman]

You have to split the ownership if you want 12 disks used for each node.

You can have an odd number of disks in a raid group, that's no problem. Whether you have 1 or 2 spares is a personal choice. One spare would give you more usable capacity. If you know you could get a replacement disk quickly - then 1 spare should be fine. If it might take awhile to get a replacement on site, you may want to have more than 1.

[heightsnj]

Thanks!

A follow-up.

When one failed, can SSD spare on the other node substitue the failed one?

[mikkisse]

Hello!

Also, if you  run on DataONTAP 8.3 you can look at storage pool technology. 

https://library.netapp.com/ecmdocs/ECMP1636022/html/GUID-8284CB30-5BC5-4D10-B0AA-AA8F8DAA752E.html

[heightsnj]

Thanks!

Could somebody please let me know if a disk failed on one node, and a spare disk on the other node could be selected for rebuilt? both nodes in the same HA.

[mikkisse]

For a disk to be used as a hot spare for another disk, it must conform to the following criteria: 

• It must be either an exact match for the disk it is replacing or an appropriate alternative.

• If SyncMirror is in use, the spare must be in the same pool as the disk it is replacing.

• The spare must be owned by the same system as the disk it is replacing.

[heightsnj]

What is your suggestion to use either Storage Pool or Dedicated SSD aggr for Flash Pool?

Thanks!

[J-L-B]

I would split the shelf - odd #'s to one controller and even #'s to the other controller.  Since the SSDs are so expensive, we chose to go with #disks+2DP with 1 spare.  The rebuild times are extremely fast on the SSD raidgroup.  You would essentially have to lose 3 disks within 4 hours if you only have 4 hour support.  Never had a SSD fail in my experience.