Remove node from master server cluster

[DaveM16]

Hi

I am attempting to remove one node from a master server SFWHA (VCS) 2 node cluster. This is as it has been decided that the second node is no longer required (a single node is sufficient).

I realise it is not possible to convert from a cluster master to non-clustered, but it should be possible to remove a node from the cluster.

What I have tried was the following

Use the cluster configuration tool to remove the node from the cluster (this worked, but the second removed node still appears known to NetBackup in EMM). At that satge attempting to use the command "nbemmcmd -deletehost -machinename <master2> -machinetype master" failed.

So I added the node back into the cluster using the same cluster configuration tool. So far back to a 2 node cluster.

Next I am trying to use the same nbemmcmd as above to remove the node from NetBackup before removing from the cluster.

This time the command fails with the following error:

C:\Users\dmoline>nbemmcmd -deletehost -machinename master2 -machinetype master
NBEMMCMD, Version: 7.6.0.1
The function returned the following failure status: "Please Delete Disk Objects First"

So I removed the all the disk STUs (I read this in another similar unresolved post), and tried again.

Now I am getting the following error:

C:\Users\dmoline\Documents>nbemmcmd -deletehost -machinename master2 -machinetype master
NBEMMCMD, Version: 7.6.0.1
The function returned the following failure status:
generic EMM SQL error (193)
Command did not complete successfully.

I'm wondering if anyone has any other suggestions.

[Marianne]

Please show us output of 'nbemmcmd -listhosts'?

You may have to remove 'cluster' entry first:

nbemmcmd -deletehost -machinename master2 -machinetype cluster

[DaveM16]

Hi Marianne

Not sure that would work - the cluster entry is associated with the cluster node name (and I'm not trying to detstroy the cluster per se, just turn in into a single node cluster).

Here's the output requested:

C:\Users\dmoline>nbemmcmd -listhosts
NBEMMCMD, Version: 7.6.0.1
The following hosts were found:
server           masterclst
cluster          masterclst
master           master1
media            media3
ndmp             media3
remote_master    remmaster1
replication_host remmedia3
master           master2
virtual_machine  vcenter1
Command completed successfully.

[Marianne]

I thought there would be a cluster association with the node names as well.... Sorry...

I remembered something in the meantime - in pre-7.0, adding a node to a cluster was not easy and certainly not documented.

I was battling to get it working and my nbemmcmd command 'sort of' worked but not quite...

I used the same command to remove the 2nd node as the initial command, just with -deletehost.

https://www-secure.symantec.com/connect/forums/add-node-netbackup-clustered-master-server 

So, maybe adding '-clustername masterclst' may work?

If not, there may be more device-related entries associated with 2nd name.
I wonder if nbdecommission will work on a cluster node? Probably not....

*** EDIT ***

I found something else - not sure if this command/script exists on Windows and if current versions still have them:

On Unix/Linux, there are scripts to add and  remove cluster nodes:

The NetBackup cluster_add_node and cluster_drop_node performs a dual function: it updates the cluster service group and NetBackup EMM database at the same time. 

# cd /usr/openv/netbackup/bin/cluster/util

# ./cluster_drop_node -s nbu -n <drop_node_name>

[DaveM16]

Hi Marriane

So far no luck with your suggestions - but thanks for those anyway.

Adding the '-clustername masterclst' produced a syntaxt error.

The nbdecommission utility rightly fails as the server is not a media server.

The cluster_drop_node utilities do not appear to be available under windows, I did grab a copy  and looked to see what they do, which is basically what I have ben trying myself.

I will talk to support to see what the issue is.

[DaveM16 · Accepted solution]

After having support look at the problem, it was determined that there was an audit record associated with the second node of the cluster preventing its removal.

As this was/is simply a test system, and the only supported option would have been to raise a case to get backline to create a custom SQL script to address the issue, I deicided to simply rebuild the master server.

I believe what I tried and Marrianne's suggestions should have worked and would work most of the time, just not in my particular case.

Sorry I can't offer any better solution to my problem.