Quick Answer
Run Get-ClusterNode | Format-List Name, State in PowerShell, find the node that's Down or Paused, evict it, and re-add it. That clears the stale instance ID.
Why You're Seeing This
This error pops up when a node in your Windows Server Failover Cluster has a different Cluster Instance ID than the rest. That happens after a node was rebuilt, restored from backup, or the cluster database got out of sync. Maybe you had a node offline for a while and then tried to bring it back — the cluster services on that node still think they belong to a different incarnation of the cluster. The error usually shows up during validation or when you try to add a node back into the cluster. It's not a hardware issue — it's a software state problem. The culprit here is almost always a node that wasn't cleanly evicted before being rebuilt.
How to Fix It (Step by Step)
- Open PowerShell as Administrator on any node that's still healthy and can talk to the cluster.
- Check the cluster status — run
Get-ClusterNodeto see which nodes are up and which aren't. You'll see something likeNode1 UpandNode2 Down. - Evict the problematic node — the one that's throwing the mismatch. Use
Remove-ClusterNode -Name Node2 -Force. The-Forceflag is key — without it, you'll get a prompt asking if you're sure, and in a script that's a pain. - Verify it's gone — run
Get-ClusterNodeagain. The node should no longer appear. - Clear the cluster state on that node — go to the problematic node physically or via remote console. Stop the cluster service with
Stop-Service ClusSvc. Then delete the cluster database folder:Remove-Item C:\Windows\Cluster -Recurse -Force. Don't worry — this only removes the local copy of the cluster config, not the actual cluster settings. - Reboot that node — just a simple
Restart-Computerworks. The cluster service will start but the node won't join automatically since it's evicted. - Re-add the node — back on the healthy node, run
Add-ClusterNode -Name Node2 -NoStorage. The-NoStorageflag tells the cluster not to take any disks offline during the process. If you're using shared storage, you might need to remove that flag and let it mount the disks. - Run a cluster validation —
Test-Cluster -Node Node1, Node2. This ensures everything is in sync and the error is gone.
That's it. In 99% of cases, evict and re-add fixes the mismatch. I've done this on Server 2016, 2019, and 2022 — same process works across all.
If That Doesn't Work
Sometimes the eviction fails because the node is too far out of sync. Don't panic — here's the fallback:
- Force destroy the cluster on the bad node — on the problematic node, run
Stop-Service ClusSvc, thennet stop clusdiskif it exists, then delete the cluster folder as above. Then set the service toDisabledtemporarily withSet-Service ClusSvc -StartupType Disabled. Now run the cluster validation from the other node — it should pass because the bad node isn't in the cluster. - Check DNS — a node that has a stale DNS record pointing to an old IP can cause weird cluster behavior. Flush DNS on all nodes with
ipconfig /flushdnsand make sure the A records match current IPs. - Look at the cluster log for the instance ID —
Get-ClusterLog -Node Node1, Node2 -Destination C:\Temp. Search forInstanceIdin the logs. If both nodes have different values, you're hosed — the evict/re-add is your only path.
If you're still stuck after 30 minutes, I'd suspect a deeper issue, like the cluster name object (CNO) being corrupted in AD. Check the computer account for the cluster name in AD, reset it if needed, and try re-adding again.
Prevention Tips
The root cause is almost always a node that was rebuilt without cleanly leaving the cluster. So the rule is: always evict a node before you rebuild it. If you're doing disaster recovery and the node wasn't evicted, just use the evict-and-add process above — it's not a big deal.
Also, keep your cluster nodes on the same Windows build. If you mix Server 2019 and Server 2022 in the same cluster, you'll see weird errors like this more often. Check with Get-WindowsVersion on each node.
Finally, don't skip validation after any cluster change. Test-Cluster takes two minutes and saves you from a 2 AM call.