1 Copper

VNX MCx SSD Hot Sparing

Good day everybody,

I ran into an interesting conversation the other day about the new hot sparing in the VNX that I didn't necessarily have a solid confident answer for, so I thought I would bring it to the community for discussion and clarification.

The discussion was around the difference in the flash disks used for FAST Cache and FAST VP, where as FAST Cache drives can be used in a FAST VP storage pool but FAST VP flash disks can not be used as FAST Cache drives.  So in the case of this, how does one handle hot sparing.  Logic would dictate that you would place a FAST cache flash drive in the array and this would be the hot spare for both FAST Cache and FAST VP storage pools. Minimizing the number of flash disks needed to be purchased and not be used.

So if a flash disk failed in a FAST VP storage pool and the FAST cache flash disk acting as a hot spare took over for the failed drive, how now does the VNX handle a replacement drive being inserted?

  • Does the active hot spare disk then revert back to a hot spare when the replacement drive is inserted?
  • Does the active hot spare just become the drive in the FAST VP pool, then what of the replacement drive (more than likely being a flash disk for the FAST VP storage pool, not compatible to be a FAST Cache drive), do we loose hot sparing for FAST Cache?
  • Does the user have to manually initiate a drive mobility job to put the drives back in their proper locations?
  • Does the TA now have to include a hot spare specifically for FAST Cache and one specifically for FAST VP storage pools and the IE needs to configure hot spare policies?

This of course only factors in the scenario where FAST Cache drives and FAST VP drives are the same size, I would suspect that the conversation around different sized drives for these roles will be answered once the hot sparing logic is explained.

I just couldn't answer the question with certainty.   Does anyone have any experience with this from testing in the lab or the field?

Reply
3 Replies
Highlighted
4 Ruthenium

Re: VNX MCx SSD Hot Sparing

The two different families of FLASH drives: SAS/SATA FLASH (interchangeable) and SAS FLASH VP and can not be spares for each other.

https://support.emc.com/docu48786_VNX-MCx-Multicore-Everything.pdf

Figure 35 - Drive sparing matrix

There is a very good matrix in the white paper above which outlines this.  Also there are several pages dedicated to the following topics:

Hot Spares

Drive Rebuilds

Sparing Rules

Hot Spare Policy

Reply
Highlighted
5 Rhenium

Re: VNX MCx SSD Hot Sparing

If you don't have a Powerlink account, you can also try the link: http://www.emc.com/collateral/analyst-reports/h12090-emc-vnx-mcx.pdf.

Reply
Highlighted
1 Copper

Re: VNX MCx SSD Hot Sparing

Excerpt from the document that explained it well to me....

"The compatibility within each set is used for sparing, as well as for RAID group creation. VNX does not allow RAID groups to contain drives from different sets. For example, SAS and NL SAS drives cannot be used within the same RAID group.  SAS Flash VP and SAS Flash are not allowed to mix either.  However, SAS Flash and SATA Flash are 100% compatible and can be used to create a RAID group.

Pool tiers have a somewhat different approach. VNX pools support up to three tiers and every tier claims a specific RAID group set, except Extreme Performance that claims two sets.  Even though SAS Flash VPN cannot be configured in the same RAID group with SAS Flash/SATA Flash drive, private RAID groups created from each set can be a part of the Extreme Performance tier within the same pool.

In Unisphere, when drives from both Flash and Flash VP sets are present, the pool creation Extreme Performance section includes two drop-down boxes, one for each set."

With that said, it seems that it is necessary to ensure that there is at least one SAS/SATA Flash drive and SAS Flash VP drive available in the system assuming that the VNX is populated with FAST Cache drives and an Extreme Performance tier.

Thank you guys for the direction to the explanation.

Reply