A frequent topic on various EMC forums is how to layout Storage Pools according to EMC best practice, typically regarding about what Drive count to use in the RAID config.
The best practice documents* is the source of truth for this when creating pools, but what about when you expand Storage Pools? Of course the recommedation is to only expand using the same amount of disks in the existing configuration. So if you have an Extreme Performance Tier made of up 20 disks in a 4 x (4+1) RAID 5 config, you should expand by a minimum of 5 disks, or multiples thereof. This keeps all the groups balanced and performance predictable.
Normally it’s easy to determine what was used. Just highlight the Storage Pool and click Expand. The existing config will be displayed.
So here I’m shown that the existing Performance tier is using RAID 5 (4+1) and the Capacity Tier is using RAID 6 (14+2). Cool, so that’s the amounts I would need to expand by.
However what if someone, out of necessity (cost or whatever) created a Storage Pool with a non-recommended drive count ? Unisphere doesn’t tell the full story.
Here’s a Storage Pool that was created in RAID 5 with only 3 disks. Yuck, I know, but it’s just a test.
So it is suggesting that the pool should be expanded using a RAID 5 (4+1) Group, and not displaying any option to expand by the current config.
Following this guidance would lead to 2 x private RAID groups in the pool, creating an uneven ‘spread‘ of data which could very well lead to a hot-spot in the tier. No one likes this, as it make performance diagnosis and remediation more difficult than it should be.
How do we avoid this ?
- A) Use recommended drive counts when initially creating your Storage Pools. Design your config BEFORE purchase.
- B) Use NAVISECCLI to check the current config.
From a powershell session, use the storagepool -list operator, and pipe the output through a select-string Powershell cmdlet, with a context limiter **
naviseccli -h array-spa storagepool -list -id 5 -all | select-string -pattern "RAID Type" -context 1,2
Here’s the output;
So we can now see the drive counts for each tier, and make a sensible, predictabel choice about how to expand it, even in configurations that are not displayed in the Unisphere gui.
Yaaay for NAVISECCLI !
* VNX Best Practice for Performance guide VNX2 Best Practice for Performance guide
** The context switch works similarly to the unix GREP A,B and C operators to show the patterns and surrounding data when searching.