Fix ERROR_CANT_EVICT_ACTIVE_NODE (0X00001391) in Windows Cluster

Quick answer: Use PowerShell to stop the cluster service on the node, then evict it from another node with Remove-ClusterNode.

This error shows up when you're trying to evict a node that's still online and active in the cluster. Windows won't let you do that—it thinks you're about to break quorum or cause a split-brain scenario. The error code is 0X00001391, and you'll see it in Failover Cluster Manager or event logs. It happens most often when a node is still running the Cluster service and has a vote. The fix is simple: take the node out of the cluster gracefully by stopping its service first.

Step-by-Step Fix

1. Identify the node you want to evict. Open Failover Cluster Manager on any other node in the cluster. Find the node under "Nodes" in the left tree. Write down its name exactly.
2. Connect to that node via Remote Desktop or console. You need local admin rights on it.
3. Open PowerShell as Administrator on the target node. Run Stop-Service clussvc. After a few seconds, you should see the service stop.
4. Confirm the service is stopped with Get-Service clussvc. The Status should say "Stopped".
5. Go back to the other node (the one with Cluster Manager open). Right-click the node you stopped and choose "More Actions" then "Evict". If you're using PowerShell, run Remove-ClusterNode -Name <NodeName>. You should get a success message.
6. Verify the node is gone from the cluster. Refresh the Nodes list in Cluster Manager.

That's it. Once evicted, you can safely reinstall the OS or repurpose the hardware.

Alternative Fixes

If you can't RDP into the node, you can stop the service remotely.

• Remote PowerShell: From an admin node, run Invoke-Command -ComputerName <NodeName> -ScriptBlock {Stop-Service clussvc}. You need WinRM enabled and permissions.
• If the node is hung or unreachable: Force a shutdown from the hypervisor or by holding the power button. Then, on another node, run Get-ClusterNode | Where-Object {$_.State -ne 'Up'} | Remove-ClusterNode. This will succeed because the node is now down.
• Last resort: If the node won't come down and it's not the quorum owner, you can change its vote to 0. Run (Get-ClusterNode <NodeName>).NodeWeight = 0. Then try eviction again. This sometimes lets the cluster ignore the node's vote.

Prevention Tips

Never try to evict a node that's still running. Always drain roles first using Suspend-ClusterNode -Drain. Keep your cluster documentation up to date so you know which node is the witness or quorum owner. And for heaven's sake, don't mess with node weights unless you understand quorum math—you'll take the whole cluster down.