determining capacity left on vsphere installation

[umarzuki]

Hi,

Is there a tool to check on my curent setup of vsphere 4.1 whether I can fit it another VM with 4 GB RAM & 100 GB HDD and it can still get the benifit of HA?

[Dave_Mishchenko · Accepted solution]

HA focuses on CPU and memory resources.  To check for disk capacity you'll need to look at the datastore you plan to use.

To check memory it will depend on how you have the cluster setup.

1) If you're using the Percentage of cluster resources reserved option for HA, then select Summry tab (see VMware HA) for the cluster and you'll see the percentage of CPU and memory resources used.  Based on the total resources listed in the General section you'll be able to determine if you can start up your new VM.

2) If you're using the Host failures cluster tolerates setting, then vCenter will determine the number of available slots available.    Click the Advanced Runtime Info link in the VMware HA section to see the slot size details and number of slots available.

Dave
VMware Communities User Moderator

Free ESXi Essentials training / eBook offer

Now available - VMware ESXi: Planning, Implementation, and Security

Also available - vSphere Quick Start Guide

[afertmann]

HA is all about how many Host failures you want to be able to sustain.  A typical approach to this is N+1.  In this you could sustain 1 host failure.  If you did N+2 then 2 host failures.

So for instance:

You have a 3 node HA cluster and want to be able to sustain 1 host failure.  In this, you dont want to allocate more that 66% of resources on any host in this cluster.  This way if you lose a host, half of the 66% of resources used could move to 1 node left in the cluster and half could go to the other remaining node thus making 99% of resources used on each remaining host.  But ideally you want to leave around 10% overhead in resources for random spikes and other over head, lowering the resources used per host in this cluster to about 56% per host.

As for resources, which ever you have less of will be your limiting factor.  For instance you max cpu load per host is at 10% and your memory load per host is at 60%, that would be what id classify as a maxed out host.  And vice versa (10% memory usage and 60% CPU) etc.

[idle-jam]

this might be useful: http://www.yellow-bricks.com/vmware-high-availability-deepdiv/

[umarzuki]

since i'm using the "percentage of cluster resources...", set it to 10% and run DRS. It seems I have to guesstimate with failover capacities shown on the cluster's summary tab.

oh, and i forgot to mention i only have 2 hosts (1 esxi 4.1 & 1 esx 4.1) and 1 vcenter. already above 0.1 target host load standard deviation. no more new VM i guess...