ESXi

Hi, I do not currently use resource pools but often feel I should look more in to it.  On our four node 5.1 cluster we have 16 web servers (among other VM's) that are exactly the same and perform the same function.  I was wondering if those would be good candidates to add to a resource pool so they can grab resources as needed?  I tend to statically overcommit the RAM and CPU on those "just in case", but maybe that is foolish if there is a better way.  Thank you.

if you want to set.... a specific... shares.... and guaranted SLA for CPU/RAM then you can use the resource pools... so when the 16 vms need resource they will expand... and they will get the configured memory/cpu during the over provisoned time....

so its all based on the design... criticality... SLA etc...

managing resources..and its shares per VM basis is not a good way.. thats why we use resource pool

Titans99 wrote:

Hi, I do not currently use resource pools but often feel I should look more in to it.  On our four node 5.1 cluster we have 16 web servers (among other VM's) that are exactly the same and perform the same function.  I was wondering if those would be good candidates to add to a resource pool so they can grab resources as needed?  I tend to statically overcommit the RAM and CPU on those "just in case", but maybe that is foolish if there is a better way.  Thank you.

When you say you "statically overcommit", are you referring to setting memory/CPU reservations?  The term "overcommittment" in the VMware world means that you are allocating more memory/CPU to your VM's than can be backed by physical resources.  It sounded like you were talking about reservations.

Best practice is not to use individual VM reservations or limits by default, and only set them on a case by case basis if you can justify doing so.  Then, use resource pools to group workloads based on some sort of criteria (Production vs Non-Production, by Department, or whatever) and use Shares to decide how physical resources are allocated when there is contention.  You can also use a Resrouce Pool level reservation to guarantee to the pool a certain amount of resources. Try to get by without setting reservations or limits, if you can.

Sorry, when I say "overcommit" I mean I assign each web server vm 6GB RAM and one CPU even though the web server may only use 3 GB.  I am not overcommitting as it relates to hardware.

Sounds like I should be using pools so I guess I should read up on them, but any examples are welcome. Mostly I get confused on how to configure the VM when it's a pool member.

Also consider that if you have no compelling reason to prioritize resources for any of the VMs over any of the other VM's, then you really don't have a reason to use resource pools at all.  In that case, each VM will be granted whatever memory it needs on demand, DRS will balance the load among hosts if one host is out of balance, and there will be fairness when ESXi requests pages back from VMs when there is contention.

Check out the vSphere Resource Management Guide, you will find out everything you ever wanted to know about Resource Pools and any other resource management topics in general.

http://pubs.vmware.com/vsphere-51/topic/com.vmware.ICbase/PDF/vsphere-esxi-vcenter-server-51-resource-management-guide.pdf

Good point, I'll have to think about prioritization.  I am currently setting the vm RAM and CPU settings on the "Hardware" tab, but leaving the "Resources" tab as defaults ("Normal" shares, "0" reservation, "Unlimited" enabled).  When you say "each VM will be granted whatever memory it needs on demand", would I need to change that strategy (assuming I don't use resource pools)?  thank you

Yeah, I'd leave the settings on the Resources tab on the VMs at their default.  I think what you are doing now is fine.