Failover Cluster - Adding an additional Host

[broadld]

Good morning all,

 

So we have currently got 2 x physical servers with Windows Server 2022 setup in a failover cluster with a SAN.

I have created new clusters in the past and done migrations etc.

One thing I haven't done is add an extra host to an existing cluster.

One thing I am worried about is that when adding an additional host is it corrupting the shared storage where the VHD files are stored. As I have heard horror stories of this happening from forum posts. Such as this one https://straightpathsql.com/archives/2020/06/cluster-validation-beware/

I have done a lot of research on this but can't find a definitive answer.

[Chuckster]

Failover clustering of what - file servers, Hyper-V?

 

The recommendation is to have a 3-node host in order to prevent split-brain.

 

When you had another host into the cluster it will go away and do its thing without any further input from yourself.

[broadld]

Failover clustering of what - file servers, Hyper-V?

 

The recommendation is to have a 3-node host in order to prevent split-brain.

 

When you had another host into the cluster it will go away and do its thing without any further input from yourself.

 

Sorry forgot to mention. It is a failover cluster used for Hyper-V. Forgot to include that part!

[pablo007]

Sorry forgot to mention. It is a failover cluster used for Hyper-V. Forgot to include that part!

 

So you have a clustered shared volume available to both hosts currently?

[broadld]

So you have a clustered shared volume available to both hosts currently?

 

Good afternoon yes there are 3 CSV's currently in the cluster. But don't want to corrupt the data on these when adding a 3rd host.

[broadld]

Anyone else with any advice or past experience adding an additional host to an existing cluster that is in use?

[pablo007]

I really do think this is add the new host but do not run any validation checks (especially not the storage part) as this will take everything off line and as long as your SAN is mapped to your server and MPIO is configured, the drives will appear in the second host drive manager, do nothing with them just leave them as they are off line or uninitialized.

 

If you use Windows admin center and the fqdn of the cluster there is a add node wizard you should be fine. The existing csvs will then appear on the third host

Edited April 26, 2024 by pablo007

[broadld]

I really do think this is add the new host but do not run any validation checks (especially not the storage part) as this will take everything off line and as long as your SAN is mapped to your server and MPIO is configured, the drives will appear in the second host drive manager, do nothing with them just leave them as they are off line or uninitialized.

 

If you use Windows admin center and the fqdn of the cluster there is a add node wizard you should be fine. The existing csvs will then appear on the third host

 

Excellent thanks for this. My plan was to:-

 

- Install Windows on the host & run updates to the same state as existing hosts.

- Connect ISCSI drives and configure MPIO

- Add host and not use validation for storage.

[pablo007]

Excellent thanks for this. My plan was to:-

 

- Install Windows on the host & run updates to the same state as existing hosts.

- Connect ISCSI drives and configure MPIO

- Add host and not use validation for storage.

 

Sounds like a plan.

[KK20]

Failover clustering of what - file servers, Hyper-V?

 

The recommendation is to have a 3-node host in order to prevent split-brain.

 

When you had another host into the cluster it will go away and do its thing without any further input from yourself.

 

I have run two host clusters (with quorum) for years, hosts (and quorum) are even in different buildings. You can set "sites" for each and configure that if one is isolated (from host and quorum) it will gracefully shutdown leaving the other site as master. Powershell will let you bring this site up and set as master if the primary site fails AND the interlink AND the quorum fails - thus avoiding split brain.

 

It works too. We had a power failure in primary site once, I set the second site to master so that PCs in the second site could continue working, primary site came back on this was set to secondary, sync completed and I returned primary back. Downtime on secondary was about 15 minutes, primary downtime was zero once the power came back (sync'd back in an hour or so in the background).

 

I have added hosts to our cluster, I was upgrading our physical boxes and storage. This was a two step procedure as I was also moving storage. I set up new storage and nodes, joined the new nodes to the cluster, added the new CSVs (lots of redirected mode), moved the VM storage locations as appropriate, removed the old nodes.