VMAX, vSphere 5.x and Zombie Space Reclamation!

We have quite the group of customers that are currently using vSphere 5.x, VMAX and Fast VP. In previous versions of ESXi we saw some weirdness when it came to dead space reclamation, movement of data and general misbehaving in vSphere 4.1 + Thin luns + FAST VP and a variety of other variables to include VAAI, Recoverpoint SAN Tap from a VMAX to an offsite CX3-80… Like I said a lot of variables.

After looking at FAST VP with our customer for almost a year now, a multitude of conversations with the best people at EMC and probably a few conversations w/ VMware our customer who prefers to remain nameless would like to share his ‘formulas’ with the greater good. You know who you are and next time I see you I owe both you and your boss a beer!

In short, vSphere + Dead Space reclamation is wierd. How VMAX and FAST VP presents it is also very difficult to report and see what is going on. This is why I call it Zombie Space Reclamation because you don’t really know if it is dead!

FAST VP (on Enginuity 5875.198 or higher) will work with VMware 5.x datastores that are defined as a tdev and not fully allocated. It will release space from the anchored tier when it is moved to another tier, but one must understand what values DO and DO NOT change.

1) Current Subscription percent value will NOT change in the anchored tier.
2) Allocated capacity will change in the anchored tier.
3) Total written capacity will NOT match Allocated capacity on the anchored tier.
4) Allocated capacity will be correct on the destination tier.

To correctly determine if FAST VP is working properly and if you have space available in the anchored tier (if you don’t want to oversubscribe), one can use the following formulas:

A person could use this formula to validate FAST VP is releasing space from the anchored tier, if Allocated Capacity equals what the VMAX is saying.
(Capacity of Anchored tier * Current Subscription % of Anchored tier) – Allocated Capacity of Destination tier related to devices using FAST VP and not anchored in the destination tier = Allocated Capacity of Anchored tier.

Capacity of Anchored tier – Allocated Capacity of Anchored tier = Available space on Anchored tier

And here is the biggest issue that we still see going on… in regards to VMAX +VMware ESXi and Storage Vmotion…

A dead space reclamation issue is only related to Storage vMotion when a virtual disk is deleted or migrated to a different datastore…and not releasing it’s space.

This is why I like to call it Zombie Space Reclamation…because you think the space is gone and possibly dead…but it still is there existing on a TDEV someplace and not being reclaimed.

Perhaps there should be a different way vMotion should be working?  Or how it reports back to the VMAX that it should be Zeros?  Time will tell.

**Disclaimer #1 VMAX FAST VP is still very much working well in customer’s environment but the strangeness in regards to how it behaves with VMware is an interesting topic.

**Disclaimer #2  This is not VNX FAST VP.


This entry was posted in Greater Good, VMAX, vmware, vsphere. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *