VCF Stretched Clusters & DRS Resource Pools

[sevenlogic]

Hi all, I have a question regarding VCF stretched clusters and DRS resource pools.

From my understanding, Resource Pools ensure that CPU and Memory requirements are met during times of contention. And Resource Pools are defined on a cluster level, and they definitely should never be used as VM folders.

So lets say I have a VCF setup and I wanted to create resource pools to protect the NSX-T Edge Nodes, Management WLD components and a few "Tier0" database applications. The Tier1 RP is for future use (if needed).

We are not constrained on CPU or Memory in the VI WLD, but of course if we loose an entire datacenter this is where resources might start to contend with each other.

The NSX-T Edges in the Management domain are not as important as vCenter and other components, as they only power the vRA components.

My question is related to the share value (High, Low, Normal), does the strategy I am using above make sense? And also is it a best practice to make resource reservations on such components? It seems that these are only really important when there are resource constraints, but of course loose an entire datacenter in a stretched cluster setup would place such constraints on the system.

Any thoughts and feedback is greatly appreciated 🙂 Thanks!

[depping · Accepted solution]

I would make sure to use "scalable shares", as described by Frank in this blog, as that ensures the RPs get the share count they need relative to the VM count in the resource pool. https://frankdenneman.nl/2020/05/27/vsphere-7-drs-scalable-shares-deep-dive/

I also think it makes sense to configure it if there's a risk of contention when there's a full site failure. I am not a huge fan of reservations (or limits) as that makes things complex rather fast if you ask me. using shares works great for 999 out of a 1000 cases.