getAllVMsOfCluster slow vs Get - VM

Hi all

I'm trying to convert a simple script PowerCLI broadband vCO (or shares in the future, when debugging is available for them...)

The PowerCLI script is very simple, it loops through all the virtual machines in a cluster and the vRAM output average / vCPU per VM.

The problem is that the action getAllVMsOfCluster take a lot of ugly of time (more than 3 minutes!) to return on a cluster with VM ~ 200, while the Get - VM cmdlet on the same cluster take 10-20 sec.

I think the Get - VM cmdlet uses a view and the getAllVMsOfCluster Action simply traverses the objects returned by the sdk/mob...

Is there a way to speed up the process (except using the plug-in, PowerShell, which I try to avoid as much as possible).

Here is the PowerShell and vCO of this simple script version:

Here's the script PowerCLI:

$clusterName = "mycluster.

$cluster = get-Cluster-name $clusterName

$vms = $cluster | Get - VM

$nbVMs = $vms.length

$nbvCPUs = 0

$nbvRAM = 0

foreach ($vm to $vms)

{

$nbvCPUs += $vm. NumCpu

$nbvRAM += $vm. MemoryMB

}

$nbvRAM = $nbvRAM / 1024

$avgvRAMPerVM = $nbvRAM / $nbVMs

$avgvCPUPerVM = $nbvCPUs / $nbVMs

Write-Host "nbVMs: $nbVMs".

Write-Host "nbvCPUs: $nbvCPUs".

Write-Host "nbvRAM: $nbvRAM".

Write-Host "avgvRAMPerVM: $avgvRAMPerVM".

Write-Host "avgvCPUPerVM: $avgvCPUPerVM".

I converted this to a vCO feed that looks like this:

Entry:

cluster as VC:ClusterComputeResource

Output:

avgvRAMPerVM under the number

avgvCPUPerVM under the number

Flow:

Action: getAllVMsOfCluster

Entry: Entry flow "cluster."

Output stream: attribute "vms" in the table of VC:VirtualMachine

Scriptable task: "do the math".

Entry: Flow attribute "vms" in the table of VC:VirtualMachine

Output:

AvgvRAMPerVM output stream

AvgvCPUPerVM output stream

Script:

var nbVMs = vms.length;

var nbvCPUs = 0;

var nbvRAM = 0;

for (var i in SMV)

{

nbvCPUs += vms [i].config.hardware.numCPU;

nbvRAM += vms [i].config.hardware.memoryMB;

}

nbvRAM = nbvRAM / 1024;

avgvRAMPerVM = nbvRAM / nbVMs;

avgvCPUPerVM = nbvCPUs / nbVMs;

System.log ("nbVMs:" + nbVMs);

System.log ("nbvCPUs:" + nbvCPUs);

System.log ("nbvRAM:" + nbvRAM);

System.log ("avgvRAMPerVM:" + avgvRAMPerVM);

System.log ("avgvCPUPerVM:" + avgvCPUPerVM);

I thought about it...

Although it is not a lot of documentation for it to the vCO, by using a view I could speed things up.

Below, the variable 'cluster' is a VC:ClusterComputeResource taken as Input of the workflow element.

The final code looks like this:

Set a bunch of var to store the results:

var nbVMs = 0;

var nbvCPUs = 0;

var nbvRAM = 0;

Get ViewManager & PropertyCollector singleton

var viewManager = cluster.sdkConnection.viewManager;

var propertyCollector = cluster.sdkConnection.propertyCollector;

Create a ContainerView for virtual machines from this group

var vmList = new Array();

vmList.push ('VirtualMachine');

Bird = viewManager.createContainerView (cluster, vmList, true);

Create an ObjectSpec to define the origin point of the search and ignore the element root (cluster) of the research

var objSpec = new VcObjectSpec();

objSpec.obj = myView.reference;

objSpec.skip = true;

Create a TraversalSpec to identify the type of object that we are interested in.

var traversalSpec = new VcTraversalSpec();

traversalSpec.name = "traverseEntities";

traversalSpec.path = "display";

traversalSpec.skip = false;

traversalSpec.type = "ContainerView;

Attach the TransversalSpec to our ObjectSpec search root.

objSpec.selectSet = new Array (traversalSpec);

Create a PropertySpec to identify the properties that we are looking for in the selected object

propertySpec var = new VcPropertySpec();

propertySpec.type = 'VirtualMachine ';

propertySpec.pathSet = new Array ('config.hardware.memoryMB', 'config.hardware.numCPU');

Create a PropertyFilterSpec to connect the object search & property search spec

var propertyFilterSpec = new VcPropertyFilterSpec();

propertyFilterSpec.objectSet = new Array (objSpec);

propertyFilterSpec.propSet = new Array (propertySpec);

Create an object of RetrieveOptions by default (the default values are fine)

var retrieveOptions = new VcRetrieveOptions();

var result = null;

to loop through the results:

{

If (result is nothing)

{

Get the first batch of results (first 100 results by default?)

result = propertyCollector.retrievePropertiesEx ([propertyFilterSpec], retrieveOptions);

}

on the other

{

During subsequent calls, get the next batch of 100 results

result = propertyCollector.continueRetrievePropertiesEx (result.token);

}

If (result! = null)

{

for (var i in result.objects)

{

(var j in result.objects [i] .propSet)

{

var name = result.objects [i] .propSet [j] .name;

var val = result.objects [i] .propSet [j] .val;

If (name is 'config')

{

nbvCPUs += result.objects [i] [j] .propSet.val.hardware.numCPU;

nbvRAM += result.objects [i] [j] .propSet.val.hardware.memoryMB;

}

nbVMs ++;

}

While (result.token! = null); the "token" property to determine whether or not there are more results to recover beyond the current batch.

I want the vRAM GB not MB.

nbvRAM = nbvRAM / 1024;

Average vRAM & vCPU per VM in this cluster

var avgvRAMPerVM = nbvRAM / nbVMs;

var avgvCPUPerVM = nbvCPUs / nbVMs;

Save all the outputs to verify that everything works: