Under normal operation, a VERITAS Cluster Server node is removed from the cluster configuration when the Cluster Server software is uninstalled from the node using the Add/Remove Programs applet in Control Panel.  However, if this method of removal has not been used (for example, if the node has become permanently unavailable), the cluster node will remain in the configuration though the node no longer exists, and it must be removed manually.

To manually remove a node from the cluster configuration, take the following steps:

1.  Save the configuration. From a command line, type:
haconf -dump -makero

This command will "dump" the running configuration to disk and set the configuration to read-only.

2.  Stop the cluster. From a command prompt, type:
hastop -all -force

This will stop the high availability daemon (HAD) on all cluster nodes while keeping all clustered applications online and monitored.

3.  Back up the configuration. From a command prompt, type:
hagetcf -conf

This command backs up all Cluster Server configuration files to the %TEMPDIR%\hagetcf\conf directory.

4.  Stop VCS communication
Stop the "VERITAS VCSComm Startup" service on each node in the cluster, as shown in Figure 1 below.  This can be done locally through Control Panel > Administrative Tools > Services (Windows 2000) or Control Panel > Services (Windows NT) on the local node.  The service can be stopped on remote nodes by accessing the Services applet of the node through Server Manager (Windows NT) or Computer Management (Windows 2000).

Figure 1
 

5.  Remove node from configuration file
On any remaining node of the cluster, open the file %VCS_HOME%\conf\config\main.cf using Microsoft WordPad or Notepad

Locate the section starting "system [NODENAME]", where [NODENAME] is the name of the node to be removed, and delete the entire section related to that node.  Figure 2 below shows the section in main.cf that should be removed for the node called NTVCS4.

Figure 2
 

Parse the remainder of the main.cf file and remove all instances of the node name.  Node names are listed in the SystemList and AutoStartList entries under each service group name, as shown in the example below:

group ClusterService (
SystemList = { NTVCS3 = 0, NTVCS4 = 1, NTVCS2 = 2 }
AutoStartList = { NTVCS3, NTVCS4 }

6.  Verify configuration file syntax
Ensure that the syntax of the main.cf file is correct and that commas and spaces are included correctly, then save and close the file.  The syntax can be verified by running a command prompt from within the Cluster Server configuration directory and typing:
hacf -verify (Figure 3)

A blank response indicates correct syntax.

Figure 3
 

7.  Copy the configuration to remaining cluster nodes
Rename main.cf on all other cluster nodes and copy the updated main.cf to the Cluster Server configuration directory of each node.  This is illustrated in Figure 4.

Figure 4
 

8.  Edit LLTHOSTS
On each node in the cluster, open the file C:\Program Files\VERITAS\comms\llt\llthosts.txt

Delete the entry for the node, as shown in Figure 5.  Save and close the file.
Note: This must be carried out on each node of the cluster.

Figure 5
 

9.  Edit VCSPDC.CONF
From the Primary Domain Controller (Windows NT) or the PDC Emulator (Windows 2000), open the file C:\Program Files\VERITAS\vcs_pdc\VCSPDC.conf using Microsoft WordPad or Notepad

Delete the entry for the node, as shown in Figure 6.  Save and close the file.

Figure 6
 

10.  Start Cluster Server communication
Start the "VERITAS VCSComm Startup"service on each cluster node.  This can be done locally through Control Panel > Administrative Tools > Services (Windows 2000) or Control Panel > Services (Windows NT) on the local node.  The service can be started on remote nodes by accessing the Services applet of the node through Server Manager (Windows NT) or Computer Management (Windows 2000).

11.  Manually seed the cluster. This will allow the cluster to start with just one node active.  This can be done from a command prompt by typing:
gabconfig -c -x

12.  Start cluster on local node.
Start HAD on the local node.  This can be done from a command prompt by typing:
hastart

It can also be done by starting the "VERITAS High availability engine" service through Control Panel > Administrative Tools > Services (Windows 2000) or Control Panel > Services (Windows NT) as shown in Figure 7.

Figure 7
 

13.  Check that the cluster is running. From a command prompt on the local node, type:
hastatus -sum OR hasys -state

This is shown in Figure 8.  The node on which HAD has been started should show a RUNNING state.

Figure 8
 

14.  Start cluster on all other nodes
Start HAD on all other cluster nodes.  This can be done by opening the Services applet of the remote node through Server Manager (Windows NT) or Computer Management (Windows 2000) and starting the "VERITAS High availability engine" service.

15.  Review the configuration
To confirm the successful removal of the node, open Cluster Manager and review the configuration to verify that the node no longer exists.