VSAN policies - Reducing stripe width

[SayNo2HyperV]

Playing with 2 node new VSAN7 lab.  Confused with object placement / policy compliance.

Had few VMs set at SW=2 (4 capacity disks per host)

Policies applied just fine and immediately rebuilt VM components with extra stripes.

Now I am trying to reduce stripe width.  But despite policy SW=1 the VM objects won't rebuild.  Policy shows compliant yet if I view object placement I still see RAID 0 stripes.  So its not compliant.

My test VMs are below 255GB.  So FFT=1 + SW=1 should equal 6 total components.  Yet the VMs I had set to SW=2 VMs still have 8 components w/ RAID0 present even though the VMs now assigned/compliant with a SW=1 policy.

I also tested making cache reserve 5% to trigger rebuilt.  Which works and triggers rebuild back to SW=1. ( 6 components.)

Changing the policy which just differs in SW=2 to SW=1 doesn't appear to trigger rebuild. 

What am I missing to reduce VM stripe width back to 1 without modifying anything other than Stripe width setting?

Thank you

[depping · Accepted solution]

Stripe width is a "minimum" value. So fi you have it set to 5 and then lower it to 1 there's no need to changes as you are meeting the requirement of 1 stripe at minimum.

[SayNo2HyperV]

Thanks for clarifying.  After playing a bit more last night I deduced that its just a minimum setting.

But it still seems out of compliance IMO.  Stripe width in my view is - Use set # of spindles..

Lets say you've had 5 stripes on handful of VMs then introduced more VMs.  Now stripes are not helping - disks maxed IO.  Large stripe size causing more issues than benefit.

So reduce policy width to 1 on VMs, but still find your VMs IO are still hitting 5 disks?

What I discovered is if I temp change a VM to a cache reserve policy of above 0 and SW=1, it will trigger rebuild, While VM is rebuilding I set to VM to different policy of Cache=0 , SW=1, and the VM continues to rebuilt back to 1 stripe without cache reserve.  Meaning don't have to wait for cache reserve policy to complete rebuilt before setting to back to SW=1 cache=0 policy.