Measuring Replication Health in a cluster

First published on TECHNET on Dec 30, 2013

As part of running a mini-scale run in my lab, I had to frequently monitor the replication health and also note down the replication statistics. The statistics is available by by right clicking on the VM (in the Manager or Manager) and choosing the
Replication
submenu and clicking on the
View Replication Health…
option.

Clicking on the above option, displays the replication statistics which I am looking for.

Clicking on the ‘Reset Statistics' clears the statistics collected so far and resets the start (“From time” field) time.

In a large deployment, it's not practical to right click on each VM to get the health statistics.  PowerShell cmdlets help in simplifying the task. I had two requirements:

  • Requirement #1: Get a report of the average size of the log files which were being sent during the VMs replication interval
  • Requirement #2: Snap all the VMs replication statistics to the same start time (“From time”) field and reset the statistics

Measure-VMReplication
provides the replication statistics for each of the replicating VMs. As I am only interested in the average replication size, the following cmdlet provides the required information.

Measure-VMReplication | select VMName,AvgReplSize

Like most of the other PowerShell cmdlets Measure-VMReplication takes the computer name as an input. To get the replication stats for all the VMs in the , I would need to enumerate the nodes of the cluster and pipe the output to this cmdlet. The
Get-ClusterNode
is used to get the nodes of the cluster.

We can pipe the output of each node of the cluster and the replication health of the VMs present on that node

Get-ClusterNode -Cluster $ClusterName | foreach-object {Measure-VMReplication -ComputerName $_ | Select VMName, AvgReplSize, PrimaryServerName, CurrentReplicaServerName | ft}

Requirement #1 is met, now let's look at requirement #2. To snap all the replicating VMs statistics to a common start time, I used the
Reset-VMReplicationStatistics
which takes the VMName as an input. However if Reset-VMReplicationStatistics is used on a non-replicating VM, the cmdlet errors out with the following error message:

Reset-VMReplicationStatistics : ‘Reset-VMReplicationStatistics' is not applicable on virtual machine ‘IOMeterBase'.The name of the is IOMeterBase and its ID is c1922e67-7a8b-4f36-a868-5174e7b6821a.At line:1 char:1+ Reset-VMReplicationStatistics -vmname IOMeterBase+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + CategoryInfo : InvalidOperation: (Microsoft.Hyper…l.VMReplication:VMReplication) [Reset-VMReplicationStatistics], VirtualizationOperationFailedException + FullyQualifiedErrorId : InvalidOperation,Microsoft.HyperV.PowerShell.Commands.ResetVMReplicationStatisticsCommand

It's a touch messy and to address the issue, we would need to isolate the replicating VMs in a given server. This can be done by querying only for those VMs whose
ReplicationMode
is set (to either Primary or ). The output of
Get-VM
is shown below

PS C:> get-vm | select vmname, ReplicationMode | fl VMName : Cluster22-TPCC3ReplicationMode : Primary VMName : IOMeterBaseReplicationMode : None

Cluster22-TPCC3 is a replicating VM (Primary VM) while replication has not been enabled on IOMeterBase VM. Putting things together, to get all the replicating VMs in the cluster use the Get-VM cmdlet and filter on ReplicationMode (Primary or . You could also use the not-equal to operation get both primary and VMs)

Get-ClusterNode -Cluster $ClusterName | ForEach-Object {Get-VM -ComputerName $_ | Where-Object {$_.ReplicationMode -eq “Primary”}}

To reset the statistics, pipe the above cmdlet to Reset-VMReplicationStatistics

PS C:> Get-ClusterNode -Cluster $ClusterName | ForEach-Object {Get-VM -ComputerName $_ | Where-Object {$_.ReplicationMode -eq “Primary”} | Reset-VMReplicationStatistics}

Wasn't that a lot easier than right clicking on each VM in your cluster and clicking on the ‘Reset Statistics' button? 🙂

 

This article was originally published by Microsoft's Virtualization Blog. You can find the original article here.