SCCM Server Groups

What Are Server Groups?

Within SCCM Assets and Compliance you can create a collection containing Server Client membership. Once the 'Server Group' feature is enabled the properties of the collection will contain an additional option to declare that all members of the collection are part of a server group.  This means SCCM knows that the servers within this collection must be treated with specific logic and settings.

In the example below the collection contains three nodes of a cluster and only one node may be Software Updated at a time.

Once policy is received for the Software Update Group deployment an election is held in which one of the nodes will engage a lock state forcing the other nodes to wait until it has been released.

The UpdatesDeployment.log details the 'Lock State'. Once the node has taken the lock (Lock State: 1) the server will then run the 'Node Drain Script' and continue if the script has an exit code of 0 zero or if the script times out (600 seconds default).

After the main deployment is complete the 'Node resume script' is run and the lock state is returned to 2.

'Node Drain Script' [Untested in Production]

	<#
	Script Disclaimer: The sample scripts provided here are not supported under any Microsoft standard support program or service. All scripts are provided AS IS without warranty of any kind.
	Server Group - Drain Stop Script
	Collects current operation status of all nodes in Cluster - i.e. Paused, Running
	Checks if Localhost is Paused -- $LocalNodeStatus=$True
	Checks if additional nodes in cluster are paused $AdditionalNodeStatus=$True
	If $LocalNodeStatus=$False and $AdditionalNodeStatus=False Drain Stop the local node
	Checks that the local node is in Paused State
	Check that all the VM have evacuated the node
	If the $LocalNodeStatus=$False or $LocalVMNodeStatus=$False implying the drain stop has not been successful report failure
	Issue exit code 1 and revert the node to Running
	Author:
	Twitter @Syswow64blog
	Web: systemcenterblog.co.uk
	#>
	#//Initialize
	$LocalNodeIsPaused = $False
	$LocalVMNodeStatus = $False
	$OtherNodesOffline = @()
	$LocalNode = $env:computername
	#//Setting the log location
	$env:LogPath = "$env:ProgramData\SystemCenterBlog\Logs"
	#// Checking the log path
	if((Test-Path $env:logpath) -eq $False)
	{
	New-Item -path $env:LogPath -type directory
	}
	#Starting the log file
	Start-Transcript -path "$env:LogPath\PatchDrain.log" -append
	#// SMTP Config
	$mail = 'mail.SystemCenterBlog.co.uk'
	$smtp = New-Object Net.Mail.SmtpClient($mail)
	$from = 'SCCMReports@SystemCenterBlog.co.uk'
	$to = 'Jonathan.Proctor@SystemCenterBlog.co.uk'
	$sub = "Patch Start Script on [$LocalNode] $Node"
	#// Functions
	Function SendMail($subject,$message,[switch] $fail = $False) {
	Write-Output "$subject`: $message"
	$smtp.Send($from,$to,$subject,$message)
	if ($Fail) {Throw $message}
	}
	Function Check-VMNodeStatus{
	Param($NodeName)
	Process
	{
	$count = 0
	Do
	{
	write-host "$NodeName Checking Node has evacuated all running VM's.............. $(get-date)"
	start-sleep 5
	[void](Suspend-ClusterNode -drain -wait) # Graceful VM migration
	#[void](Suspend-ClusterNode -drain -forcedrain -wait) # Forceful VM migration
	$count += 1
	}
	Until (((get-VM).count -eq 0) -or ($count -eq 3))
	if ($count -eq 3){return $False}
	else {return $True}
	}
	}
	Function Pause-LocalNode{
	Param($NodeName)
	Process
	{
	$count = 0
	Write-host "$NodeName Drain node ...................... $(get-date)"
	Suspend-ClusterNode -drain -wait
	#Check that local node has been Paused
	Do
	{
	write-host "$NodeName Checking Node is in a Paused state........................ $(get-date)"
	start-sleep 2
	$count += 1
	#$PausedClusterNodes = Get-WmiObject -Class "MSCluster_Node" -Namespace "root\MSCluster" | Where-Object {$_.state -eq 2}
	}
	Until ((Get-WmiObject -Class "MSCluster_Node" -Namespace "root\MSCluster" | Where-Object {$_.state -eq 2}).name -eq $NodeName -or $count -eq 60)
	if ($count -eq 20)
	{
	write-host "$NodeName did not enter a Paused state ........................................... $(get-date)"
	return $False
	}
	else
	{
	write-host "$NodeName is in a Paused state ........................................... $(get-date)"
	return $True
	}
	}
	}
	Function Write-Output{
	Param($Node,$OperationStatus,$NodePaused,$VMStatus)
	Process
	{
	write-host "$Node has Operation Status $OperationStatus ...... $(get-date)"
	write-host "$Node $ LocalNodeStatus = $NodePaused .......... $(get-date)"
	write-host "$Node $ LocalVMNodeStatus = $VMStatus ............ $(get-date)"
	}
	}
	#Get Cluster Node Status
	$ClusterNodes = Get-WmiObject -Class "MSCluster_Node" -Namespace "root\MSCluster"
	foreach ($CNode in $ClusterNodes)
	{
	if($CNode.State -eq 2) # Paused state
	{
	if ($CNode.name -eq $LocalNode) # check if local node is paused
	{
	$LocalNodeIsPaused=$True
	write-host "Local node in Paused state = $LocalNodeIsPaused -- "$CNode.name" .... $(get-date)"
	}
	else # another node is paused
	{
	$OtherNodesOffline += $CNode.name + "," + $CNode.state
	write-host "Additional node in Paused State Will not proceed-- "$CNode.name" .... $(get-date)"
	}
	}
	elseif ($CNode.State -eq 1) # Check for Failed state and exit
	{
	write-host "Node is in failed state = $CNode.name .... $(get-date)"
	$OtherNodesOffline += $CNode.name + "," + $CNode.state
	}
	}
	if ($OtherNodesOffline.count -gt 0)
	{
	if ($LocalNodeIsPaused) # Resume the local node if this is paused also as the SCCM script is going to fail
	{
	Resume-ClusterNode
	}
	foreach ($Node in $OtherNodesOffline)
	{
	$NodeName = $Node.split(",")[0]
	$NodeState = $Node.split(",")[1]
	write-host ""$NodeName" has Operation Status $NodeState .... $(get-date)"
	}
	write-host "SCCM Exit code 1................................................$(get-date)"
	Return 1
	}
	else
	{
	if ($LocalNodeIsPaused -eq $False) # No nodes are paused, pause the local node
	{
	$LocalNodeIsPaused = Pause-LocalNode -NodeName $LocalNode # try to pause local node
	}
	if($LocalNodeIsPaused -eq $True) # Only Local Node paused, check VM Count
	{
	$CheckVMNode = Check-VMNodeStatus -NodeName $LocalNode
	Write-Output -Node $LocalNode -OperationStatus 2 -NodePaused $LocalNodeIsPaused -VMStatus $CheckVMNode
	if ($CheckVMNode -eq $true) # VM Status is true
	{
	write-host "SCCM Exit code 0................................................$(get-date)"
	Return 0
	}
	else # VM Status is false
	{
	write-host "SCCM Exit code 1................................................$(get-date)"
	Return 1
	}
	}
	else
	{
	$NodeStatus = Get-WmiObject -Class "MSCluster_Node" -Namespace "root\MSCluster" | Where-Object {$_.name -like $LocalNode}
	$CheckVMNode = Check-VMNodeStatus -NodeName $LocalNode
	Write-Output -Node $LocalNode -OperationStatus $NodeStatus.state -NodePaused $LocalNodeIsPaused -VMStatus $CheckVMNode
	Write-Host "SCCM Exit code 1................................................$(get-date)"
	Return 1
	}
	}

view raw SCCMServerGroupDrainScript.ps1 hosted with ❤ by GitHub

'Node resume script'
Resume-ClusterNode

Lock States
There are four lock states:

Waiting for Lock: 0
Have Lock: 1
Released Lock: 2
Node Script Failed: 3

The SQL script below will report on all Server Groups with Pre and Post Actions as well as current client Lock states.

	/* Script Disclaimer: The sample scripts provided here are not supported under any Microsoft standard support program or service.
	All scripts are provided AS IS without warranty of any kind.
	*/
	Select
	c.Name [CollectionName],
	LastModificationTime,
	UseCluster,
	UseClusterPercentage,
	PreAction,
	PostAction,
	ClusterCount,
	ClusterPercentage,
	ClusterTimeout
	From
	vSMS_CollectionSettings cs join v_Collection c on cs.CollectionID = c.CollID
	Where
	UseCluster = 1
	Select
	c.Name [CollectionName], s.Name0 [SystemName], dm.[State], dm.[TimeStamp]
	From DeploymentMutex dm
	join v_R_System s ON s.ResourceID = dm.ResourceID
	join v_Collection c ON c.CollID = dm.CollectionID
	Order By c.Name, dm.[State] Desc
	/*
	Lock States
	There are four lock states:
	Waiting for Lock: 0
	Have Lock: 1
	Released Lock: 2
	Node Script Failed: 3
	--
	*/

view raw SCCMServerGroups.sql hosted with ❤ by GitHub

gg