Windows server vmdk files spread across multiple datastores

[vozdra]

I'm very new to VMware so I have one question. What is your opinion about speeding vm Windows 2016 server vmdk files across multiple datastores? Is this good practice or it is better to have all files on one datastore? I also have Veeam backup software so I’m little concern about backup and restore in vmdk files across multiple datastores scenario.

[daphnissov]

Generally speaking, it's not a great idea because it extends the blast radius of datastore failures and it makes it more difficult to troubleshoot performance problems should they arise.

[continuum]

> I also have Veeam backup software so I’m little concern about ...

Veeam is not a tool that you switch on and forget.
As soon as you have created an automatic backupjob you have a larger risk to lose that VM than you had before.
A VM using using eagerzeroed vmdks only - ideally with mapping-details stored outside of the VMFS is something that will run foine even if you do not look at it for half a year.
When you launch Veeam you can expect that it fails to clean up snapshots once every half year or so.

So you cooperate  with Veeam and :
- setup your VMs as simple as possible: all vmdks in the same directory
- used thin basedisks
- check your VMs for errors with MS-shadowcopy providers
- do not use hot snapshots at all and only add cold snapshots when it is really necessary

- and most important - read Veeam-logs every morning - do not let snapshots pile up !!!

If both parties do their part of the teamwork - you will sleep better

 

Ulli

I

[a_p_]

>>> Is this good practice or it is better to have all files on one datastore?
It actually depends on the VMs workload. For general purpose VMs, you'd likely not see any benefit. However for special use cases like high performance databases, and depending on your storage, it may make sense to place the OS, the database, and the logs on separate datastores on different storage tiers.

André