What is vsan Inaccessible Capacity?

[Luca82]

Hi everyone, firstly I would to share this image:

[img]http://imagehosting.biz/images/2018/02/28/2018-02-2817_05_45-vSANHybridTCOandSizingCalculator.png[/i...

this is the result of a sizing scenario. I have 15vm with 80GB HDD each for a total of 1200GB. With a view to guarantee FTT1 I consider 1200GB * 2 = 2400GB....ok add some disk for vsan architecture and vswp file etc.... but here I see additional free space for slack (what is slack) and moreover lot of "Inaccessible Capacity" after an host failure... What is and why it is inaccessible?

Thank you everyone.

[Luca82]

the image:

[TheBobkin]

Hello @Luca82,

Slack is the unused space on vSAN that is advised left as 'slack' for reconfiguring Objects (as this uses space temporarily) and for rebuilding components lost due to disk/host failures. VMware recommendation for slack is ~30-25% of total vsandatastore capacity.

The Inaccessible capacity is capacity of the vsandatastore that would not be accessible if one node failed - if a host has 2TB storage locally-attached and contributing to the vsandatastore and that hosts PSODs/gets isolated then this 2TB is not available.

Bob

[Luca82]

Ok thank you. Now I still have some trouble to understand how to do sizing. For example, using the vsan ready-node tool, I've filled the workload for 15vm (1200GB total space) and the result is 8 ESX server (they seems too many!!)

Is it possible to forced to configure a 2+2 stretched cluster so 4 host total (2 per site)? Or maybe 1+1 (si physical host) plus witness?

I've tryed to use the other sizer tool... and forced to configure a robo infrastructure... this is the result

I like it. I can assume that I need to add the witness.... do you think could it be ok???

Thank you again

[TheBobkin]

Hello Luca82​,

The vSAN ReadyNode Sizer appears to be incapable of imaganing clusters smaller than 4-nodes, it also appears to just double the current advised inventory if you select stretched (!!!) - I will aim to contact the developer of this tool as this is not fit for purpose.

Your capacity usage and VM count is very low so unless these are monster-VMs I wouldn't think you would require more than a 1+1+1 configuration(or 2+1 if going DirectConnect)

Are you planning on implementing a streched cluster here?

If so then the 2-node layout in your last screenshot there should do the job - though switch H730P controllers for HBA330 if possible as these work better with vSAN (and are similar price).

If you want to (likely) improve the performance of this and/or allow room for growth, consider 400GB cache-tier drives instead of 200GB.

"I can assume that I need to add the witness"

If you are making a stretched or 2-node cluster, yes a Witness running at a 3rd site is required.

Bob

[Luca82]

Thank you again Bob, just another clarification... I've noticed that there are several sizing portal.... for example, the following is fot hybrid and all flash

https://vsansizer.vmware.com/

and this one just for hybrid

https://vsantco.vmware.com/vsan/SI/SIEV

but in both portal I can size hybrid.... the problem: if I input the same values...they show me different configurations!!!!! Which one is to be cosidered as the better choice?

Only with the second link is possible to use the "ROBO" configuration (wich I use to simulate the stretched cluster)....

But this isn't over! If I go to the vsan ready node configurator

http://vsanreadynode.vmware.com/RN/RN

I noticed that there is not matches with the advices of other links. Just an example:

15vm

80GB HDD each

4GB RAM each

the vsan sizer suggests me the HY-2 profile with 24cores and 256GB RAM....but there aren't any ready node with that configurations! I can't understand....

can someone help me to better understand?