Over Provisioning policy

[cryton2707]

Hi all,

Ive got some 300 guests that Ive thin provisioned.  Now Ive got loads of vmfs that are over provisioned..

Basically people requesting VM's with much larger disks than the data they actually have in the end.

And It's usually fairly static.. But I'm looking at this and thinking this could become a potential issue.

So Im wondering what other people do as to reguarding by how much they allow a disk to be over provsioned by.

Bascially I need to create a policy that says once the disk get to x amount over provisioned no more vm's go on that

disk. or x number of vm's per vmfs..

I just looking for some info on what other people have done.. so I can then tackle our over provisiong.. Before it becomes

major issue.. 

Thanks All

:smileylaugh:

[chriswahl]

My advice is to be "tough but fair" on allocating disk space per user requirements. Often times requests come in for a "50 GB data drive" because 50 is a nice, round number. Question further as to why the amount of space is needed and explain that disk space is not an unlimited resource. Make sure your management team is on board with this, and explain what happens when a LUN runs out of space and the effects on the VMs that use that LUN. That said, I don't think there is a boiler plate "120%" or some other number, it really depends on the environment and what the VMs are running - having 300 web servers might be less of a worry than 300 application and database servers.

Personally, the only place I consider to be "comfortable" to get by with overcommitment is on the OS disk LUNs when there is storage deduplication / compression on the SAN. By arranging LUNs by OS (2003, 2008, etc.) you can usually achieve ridiculous amounts of deduplication as all of the LUNs contain similar OS data.

[msemon1]

I have many Thin Provisioned VM's in my Test and development environment. This recovers alot of wasted space on the SAN. I do this for things like templates as well. If you are going to do this and are worried about running out of SAN space, set an alarm for something like 80-90% for disk space on the SAN so you can move things around before you run out of space.