Usable Storage x Raw Storage

[sandroalvesbras]

Hi,

I found these and views that shows when space I am using and how much I have available.

However, the only view that shows when I actually have it is the latest version of vSAN 6.7.

In previous versions, how do I see this free space that shows in the latest version of vSAN?

Thank you.

vSAN 6.5

vSAN 6.7

[TheBobkin]

Hello sandroalvesbrasil​,

"In previous versions, how do I see this free space that shows in the latest version of vSAN?"

Capacity overview has been present since 6.0 U2 in the same place you found it in modern versions (FLEX Client):

Cluster > Monitor > vSAN > Capacity

Older versions won't show the 'Free Usable with Policy' though as this feature was only implemented later.

Bob

[sandroalvesbras]

Was this information available in which version?

Can I view this information or otherwise infer using the previous version?

I have an environment with vSAN 6.5 and it does not show.

[TheBobkin]

Hello sandroalvesbrasil​,

"Can I view this information or otherwise infer using the previous version?"

Which information, 'Free Usable with Policy'? If so, then no as this was not present prior to 6.7 U1.

You can of course work this out by getting datastore size minus slack-space and then divide by 2 (assuming using FTT=1, RAID1).

Bob

[sandroalvesbras]

Hi,

really....

I created a lab with vSAN 6.7 U1 and saw in practice how it calculates.

Just get the value of Free RAW that appears in the capacity, divide by two and decrease by 30% that I have the recommended free value.

I also noticed that when I choose the datastore of vSAN to create the VM it does not appear the option to create a Thick disk. That is, the policy that defines that in this case all VMs are Thin.

But I had another question:

- That means I can create countless, but if we do not have control of the size of the discs based on the total usable space, I can cause an overbooking ...

Or will vSAN warn me that I'm causing overbooking?

Thank you.

NOTE: Ignore the execel worksheet information. Has information from another vSAN cluster.

[TheBobkin]

Hello sandroalvesbrasil​,

"Just get the value of Free RAW that appears in the capacity, divide by two and decrease by 30% that I have the recommended free value."

This is literally exactly what I informed you of in your other (near identical) post.

"- That means I can create countless, but if we do not have control of the size of the discs based on the total usable space, I can cause an overbooking ..."

If you create 100 thin-provisioned VMs with initially 1GB used each and disk-size of 1TB then you can of course run out of space due to the space being consumed as the VMs grow - this aspect is no different from any other VMFS-based datastore running on a SAN.

"Or will vSAN warn me that I'm causing overbooking?"

You will get alarms and warnings when you start reaching high consumption points on disks (yellow warning at 80%, red alert at 95%) as I also said in other post.

vSAN won't tell you what your over-provisioning ratio is (e.g. 200% utilisation with all vmdks full) - if you are going to thin-provision then plan accordingly as is required in vSAN and non-vSAN environments alike.

Bob

[sandroalvesbras]

These alarms you quote, are for occupation of the disks individually they are part of the disk group by node?

[TheBobkin]

Hello sandroalvesbrasil​,

There are Health UI based alerts (e.g. for individual disk usage as I said above) and standard vCenter-based alarms (e.g. for vsanDatastore overall usage) which can be custom configured however you want:

vSAN Capacity Management and Monitoring Part 2 - Virtual Blocks

Bob

[sandroalvesbras]

Hi,

I read this documentation before, but I have great difficulty understanding what the benchmark usage rate these percentages refer to.

It is quoted in the explanation that when disk reaches 80%, then it refers to the percentage of disk capacity without considering RAID-1, but rather on the basis of RAW usage.

I would like something related to usable free space to generate alarms and let me know in advance.

Tks.

[TheBobkin]

Hello sandroalvesbrasil​

"It is quoted in the explanation that when disk reaches 80%, then it refers to the percentage of disk capacity without considering RAID-1, but rather on the basis of RAW usage."

So, in reality you can have shortage of space at 3 levels:

- vmdk running out of space due to being full, e.g. ~100GB used within Guest-OS on a 100GB vmdk (which let's say occupies 200GB as 2x100GB replicas).

You can't set alarms for this in vSphere other than with vROPS, but of course there are 3rd-party monitoring alternatives that can achieve this such as PTRG, Veeam One etc. .

- Individual capacity-tier devices (or Disk-Groups if using dedupe) becoming full due to a rapid increase in usage are/or limited available disks with free space in which components can be written. e.g. if you had a 3-node cluster with only a single 100GB capacity-tier device available on each node and created a 90GB vmdk (FTT=1,RAID-1) this would store 90GB on Node1, 90GB on Node2 and a 16MB witness component on Node3 - so while you might say 'but I have ~120GB free on vsanDatastore' in reality you have 2 available Fault Domains at 90% full.

This is why disk fullness is important to be aware of and monitor.

In reality no-one is running single disk vSAN nodes but when it comes to component placement something similar to the above may be occur at high % used levels due to the fact that thin components grow in place and other factors (e.g. small capacity-tier devices) and that vSAN can only move data off disks >80% at a given rate.

Disks becoming full should be avoided thus the recommended slack space and the improvements that vSAN has made over time to avoid this (e.g. reactive rebalance, striping components to fit in smaller spaces, transient component clean-up).

Cluster alerts generated by vSAN Health are a good method of monitoring, but as with any form of alert system you would want to either check these periodically or set-up automated alert (e.g. via email).

- vsanDatastore running out of space - more likely you will run out of usable space as noted in explanation above.

Regarding standard vCenter datastore capacity alarms, these should be tuned accordingly if wanted (e.g. warning at X%, alert at Y%, SNMP etc.) and are no different to configuring non-vSAN datastore alarms.

Bob