Admission Control Using Percentage of Cluster Resources Reserved Policy

[angelok]

Hi all,

I want to understand something:

I have a VMware HA Cluster with 2 identical nodes. Normally, I should reserve 50% of resources (CPU and memory) at least as failover capability.

This would mean that if an HA failover happens, a server will be busy at 100% (50% +50%), which could be worse for performances.

According to my logic, this means that we must also reserve a certain percentage of resource for the host to avoid that it be fully busy.

Do you have an idea of ​​the percentage values ​​or resources to be reserved for the host?, What are the VMware best practices about this issue ?

Thank you very much!

[marcelo_soares]

VM resource utilization really depends on several factors. According to your logic (which is not wrong, but very simplistic) you can user 100% of resources of a specific host (the resource is there to be used, there is no reason to waste X% of resources "just in case") in case of HA failure - you will be recovering the lost host ASAP, I imagine.

HA reserved resources are calculated in a per-slot basis. The KB http://kb.vmware.com/kb/1010594 demostrates how this is performed on the vCenter. Having this in mind, you can adapt/calculate slots for your systems, and then know what will happen in case something fails on your cluster.

Hope this helps.

[LordISP]

My experience was you have to put some reservation to the VMs when you want use the percentage model. It's correct that the cluster reserve some resources for a failover scenario but the commitment stays still up to you which neans in case of a host failure HA may restart all failed VMs but you'll have less resources available.

To solve this issue you have to put reservation on VMs. Once the VMs is powered on admission control reserves the entire reserved VM capacity.

I wrote the automation script that run all five minutes, calculates the proper cluster percentage and put resource reservation to all VM that follow a specific naming convention. This allows you to different between jit, prod, dev, test etc.  environments.

Here the link: http://wp.me/p37XPc-4

Cheers

Rafael

http://rafaelcamison.wordpress.com

[depping]

I wrote about a billion articles on Admission Control by now: Search for "admission control" - Yellow Bricks

Anyway, first of all... Admission Control is not about "resources used" but rather "resource reserved". When you do not have a reservation for CPU or memory then default values are used. So even when you set "admission control" to 60% it could still be that when 1 host fail ALL resources are actively used. The VMs can always demand more than what is statically reserved for them right.

I would just monitor DRS resources and ensure you aren't running at over 60% with two nodes in your cluster. I guess the question is, how much of a performance hit can you take when there is hardware outage.

[LordISP]

Yeah that's also not wong but you have more control about your resources with the method I recently posted becaus you decide how many hosts can fail without impacting the infrastructure but still alow overcommitment