Remove node from Service Guard cluster

[David Scheiman]

Group,

I have a Service Guard cluster version 10.10 on
HPUX 10.20 that has had a node decomissioned and removed without updating the culster
configuration. I would like to clean up the
configuration.

The doc "Managing Service Guard" only mentions the scenario of removing the node from the
cluster while the node is still present.

To me the most logical way to proceed would be to remover all references to node from the cluster and package configurations. Since this is a production system, I placed a
call to HP tech support for advice. They were not sure never having tested removing a node from the cluster configureation after the node is physically removed. They suggested that I
could test the methode described above, or
seek out assistence from this forum.

I do not have a cluster available to test and
qualify this procedure, so here I am.

Thank you and kindest regards,

Dave

[A. Clay Stephenson]

Hi David:

I have not tried this scenario and frankly would have never even considered it but the 'Managing Service Guard' manual does indicate in Table 7-1 under 'Delete a Node'
indicates that 'A node can de deleted even though it is unavailable or unreachable'. That certainly seems to be you.

A few pages on, under the 'Using MC/ServiceGuard Commands to Delete Node from the Configuration While the Cluster is Running' section, it walks you through the steps but there is a NOTE about unreachable nodes indicating you may need to halt the cluster.

It looks like this is doable but if it were me, I would call Support back and ask why they didn't find the same procedure and make them give you an answer.

Regards, Clay

[Santosh Nair_1]

You should be able to remove the node from the cluster without taking the cluster down. From the "Managing MC/ServiceGuard" manual.

Using SAM to Add Nodes to the Configuration While the Cluster is Running
<>
Use the Cluster Configuration area of SAM, and choose "Modify Cluster Configuration" from the Actions menu. Then select "Modify Cluster Name and Nodes." Highlight the name of the node you wish to add to the cluster in the list on the right of the screen, then click the Add button. When you are finished changing the node list, click OK.

Using MC/ServiceGuard Commands to Add Nodes to the Configuration While the Cluster is Running
<>
Use the following procedure to add a node with HP-UX commands. For this example, nodes ftsys8 and ftsys9 are already configured in a running cluster named cluster1, and you are adding node ftsys10.

1. Use the following command to store a current copy of the existing cluster configuration in a temporary file:

# cmgetconf -C temp.ascii

2. Specify a new set of nodes to be configured and generate a template of the new configuration:

# cmquerycl -C clconfig.ascii -c cluster1 \ -n ftsys8 -n ftsys9 -n ftsys10

3. Edit the file clconfig.ascii to check the information about the new node.

4. Verify the new configuration:

# cmcheckconf -C clconfig.ascii

5. Apply the changes to the configuration and send the new binary configuration file to all cluster nodes:

# cmapplyconf -C clconfig.ascii

Use cmrunnode to start the new node, and, if desired, set the AUTOSTART_CMCLD parameter to 1 in the /etc/rc.config.d/cmcluster file to enable the new node to join the cluster automatically each time it reboots.

-Santosh

[melvyn burnard]

I would follow A.Clay's suggestion and call your HP Support office back!
They should certainly not have fobbed you off and pointed you at this site (unless you do not have support with them ?)
If the node is no longer available, you will need to remove all references to the node from the cluster ascii and package ascii files, halt the cluster for safety, and re-apply the binary with cmapplyconf.

[Christopher McCray_1]

You are correct, all you need to do is remove the references to the node in the package and cluster config files. Then do :

#cmcheckconf -v -C cluster.conf (cluster config file)
#cmapplyconf -v -C cluster.conf

You shouldn't have to halt the cluster; hp does say you should halt all packages referencing it, but they also say you can remove a node name from the package.conf file without halting the package, so use you best judgement.

Hope this helps
Chris

[Sanjay_6]

Hi David,

Look at the thread below for more info.

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90053/B3936-90053_top.html&con=/hpux/onlinedocs/B3936-90053/00/00/48-con.html&toc=/hpux/onlinedocs/B3936-90053/00/00/48-toc.html&searchterms=delete%20node&queryid=20011025-063206

Have never deleted a node from the cluster till now. But the manual says you can delete the node even if it is unreachble.

While applying the configuration you have to give the name of the existing cluster alongwith the name of the nodes which are still going to be part of the cluster and omit the name of the node, not going to be part of cluster. It advises you to halt the cluster, in case of any errors.

hope this helps.

thanks

[David Scheiman]

Thank you everyone for your replies!

From Melvyn and Christopher's responses it
looks like my original approach should work.
I feel most comfortable with modifying the
ascii configuration files by hand, so I'll give
that a try once I get a maintenance window.

BTW, we do have a high level support
aggreement, and I do believe that I should've
received a better answer. The fact is I am as
busy as the next engineer. I need to perform
my professional due dilegence, and attend to
the next emergency/disaster/priority project...

I have received a much better response here
than I have from HP in my last couple of
encounters.

Thanks again and I hope that I can return
the favor soon.

Regards,

Dave