Enable quick restarts with Kernel Soft Reboot
Applies to: Azure Stack HCI, version 21H2
New in Azure Stack HCI, version 21H2, Kernel Soft Reboot (KSR) is a premium feature available on all Azure Stack HCI integrated systems. You can use it when updating the cluster to reboot the servers faster than a normal reboot. This topic explains when to use a KSR over a normal reboot and provides instructions on using this feature on Azure Stack HCI.
Why use Kernel Soft Reboot?
Traditionally, servicing a cluster (for example, applying software updates) requires putting each server in maintenance node, evacuating the server, installing updates, rebooting if necessary, and then repeating the process for every server in the cluster. Even though Cluster-Aware Updating (CAU) automates many of these tasks, when the cluster resumes, any new data written while under maintenance mode still needs to be resynced. The longer it takes to reboot a server, the more data must be resynced, and the longer the overall cluster update time.
Kernel Soft Reboot improves reboot performance by streamlining the operating system flow, minimizing the amount of data to be resynced and therefore reducing the overall cluster update time. The amount of time saved will be proportional to the memory and size of the server. As server resources such as available memory and drives increase, so does the time savings.
Take a few minutes to watch the video comparing the performance of a normal reboot to a Kernel Soft Reboot on an idle server.
When to use Kernel Soft Reboot
As this feature bypasses the lengthy and traditional reboot BIOS/FIRMWARE initialization, you can only use Kernel Soft Reboot for updates that do not require a firmware/BIOS initialization. Currently, you can use Kernel Soft Reboot optionally with the Cluster-Aware Updating WindowsUpdate plugin for Quality Updates and Hotfix plugin for MSI/MSU/EXEs files only.
Enable Kernel Soft Reboot with Cluster-Aware Updating using PowerShell
In Azure Stack HCI, version 21H2, you can use PowerShell to set up and manage Kernel Soft Reboot for your Azure Stack HCI cluster.
The default reboot option for CAU updates is a normal reboot unless Kernel Soft Reboot is explicitly enabled using one of the following two options.
Option 1: Enable Kernel Soft Reboot for all future CAU runs
Set the cluster private property with key
CauEnableSoftReboot to value 1 if you want CAU to set Kernel Soft Reboot as the default reboot option for all supported updates.
Get-Cluster | Set-ClusterParameter -Name CauEnableSoftReboot -Value 1 -Create
Resetting this private property to any other value than 1 or deleting the property entirely will disable Kernel Soft Reboot on the cluster.
Option 2: Use Kernel Soft Reboot with individual CAU runs
To use Kernel Soft Reboot with a specific Cluster-Aware Updating run, use the optional
AttemptSoftReboot parameter when using CAU PowerShell cmdlets, such as in the following examples.
Invoke-CauRun <other_options> -AttemptSoftReboot
Add-CauClusterRole <other_options> -AttemptSoftReboot
Set-CauClusterRole <other_options> -AttemptSoftReboot
Skip Kernel Soft Reboot on certain servers
Ideally, all the servers in an Azure Stack HCI cluster should support Kernel Soft Reboot. However, you can set a registry setting on a server to make it opt out of using KSR.
Setting this registry value on any server in the cluster will cause Cluster-Aware Updating to skip Kernel Soft Reboot and attempt to reboot the server normally.
Key: SOFTWARE\Microsoft\Windows\CurrentVersion\ClusterAwareUpdating Name: CauBypassSoftBootOnNode Type: REG_SZ Value: True
Disable Kernel Soft Reboot for an update run
If you're installing operating system or hardware updates with Windows Admin Center on a cluster that has Kernel Soft Reboot enabled, you'll see a Disable Kernel Soft Reboot for this run checkbox. Checking the box disables Kernel Soft Reboot only for that particular updating run. This makes it possible to leave Kernel Soft Reboot enabled on a cluster, but disable it when an updating run requires a full reboot, such as BIOS updates.
To determine the type of reboot that was last performed, use
detailed parameters to get the report for the last Cluster-Aware Updating run.
$report = Get-CauReport <other_options> -Last -Detailed $report.ClusterResult.NodeResults | fl Node,NodeRebootResult
The report should return a
NodeResults list that contains the
NodeRebootResult for each server in the cluster. The output should look something like this:
Node : VM01 NodeRebootResult : RebootSummaryResult : Succeeded BootType : SoftBoot SoftBootStatus : Enabled
The following table shows how each name-value pair provides more information on the last reboot for each server in the cluster.
|Node||States the name of the node in the cluster.|
|RebootSummaryResult||Mentions if the last reboot was successful or not irrespective of the type of reboot requested and type of reboot performed. For example, if a user requests a KSR but a normal reboot was performed successfully, RebootSummaryResult would still return the value Succeeded.|
|BootType||Specifies the type of reboot performed in the last run.|
|SoftBootStatus||Indicates if Kernel Soft Reboot is Enabled, NotEnabled, NotInstalled, or Bypassed if the server opts out of Kernel Soft Reboot using registry key settings.|
For additional troubleshooting, see Save-CauDebugTrace.
This section answers frequently asked questions about Kernel Soft Reboot on Azure Stack HCI, version 21H2.
Can I use Kernel Soft Reboot with Windows Server?
- No. Kernel Soft Reboot is only available in Azure Stack HCI, version 21H2 and may only work on those validated as Integrated Systems.
How can I tell if my hardware supports Kernel Soft Reboot?
- All hardware marked as Integrated Systems in the Azure Stack HCI Catalog can perform Kernel Soft Reboot.
Will Kernel Soft Reboot work with Validated Nodes?
- Although this feature is not blocked on other HCI hardware, Validated Nodes are not required to test or support Kernel Soft Reboot. If you're using a Validated Node, you can try using Kernel Soft Reboot using the instructions in this article.
For more information, see also: