Adding a New node to an existing Cluster

[Andrew R.]

Hello Everyone..
I have a two node cluster sysA and sysB they are running one Package only. I need to add one more node sysC and one more package. The purpose is to have sysA and sysC failin over into sysB. My question is "Do I have to do this OFF line" or do I have to re-applyconf. Also. "Does the lock disk has to be the same on all three nodes ?" Here is the steps which I thought I should do. Please correct me If I'm wrong,
sysA>cmhaltpkg pkgA
sysA>cmhaltcl -f
sysB>cmhaltcl -f
sysA>cmdeleteconf -c clusterA
sysA>vi cmclconf.ascii 'add 3rd node info'
sysA>cmcheckconf -C cmclconf.ascii
sysA>cmapplyconf -C cmclconf.ascii ; cmruncl
sysC>cmapplyconf -P pkg_conf

Is this enough ??!! ; Thanks a lot

[A. Clay Stephenson]

Hi Andrew,

If you were only adding a node, you could do that with the cluster up. The lock disk must be accessible from all the nodes. Since you are going to a 3-node cluster, I would dispense with the cluster lock disk. I would simply remove the references to the cluster lock vg in the conf file.

[Christopher McCray_1 · Accepted solution]

A cluster lock disk is not required for a three node cluster, but is recommended, and will need to be visible by all nodes. As for the adding a node, do the following:

# cmgetconf -v -c temp.conf
#cmquerycl -n -n -n -C cluster.conf
#vi cluster.conf (use the temp.conf file to apply any other settings you want saved to your cluster config, ie. NODE_TIMEOUT, etc)
#cmcheckconf -v -C cluster.conf
#cmapplyconf -v -C cluster.conf
#cmrunnode

You don't even have to halt the cluster to do this. All you would do at ths point is add your new package and you're ready to go!

Good luck,
Chris

[linuxfan]

Hi Andrew,

You do not need to halt your existing cluster to add a new node.

http://docs.hp.com/cgi-bin/onlinedocs.py?mpn=B3936-90045&service=hpux&path=../B3936-90045/00/00/49&title=Managing%20MC/ServiceGuard

Once you add the node, you can add a new package as well.

Yes the lock disk has to be shared by all the nodes in the cluster, Also the disks in pkg1 have to be shared by sysA and sysC, and disks in pkg2 have to be shared by sysB and sysC. But there is no problem if all the machines share all the disks.


-HTH
Ramesh

[melvyn burnard]

Rather you need to do this:
1) sysA>cmhaltcl -f
2) ensure you have the discs connected to all three nodes.
3) import the shared VG's to the new node
4) vgchange -c n on the shared VG[s]
5) vgchange -a n the shared VG[s] on sysA
6) sysA>cmquerycl -v -C /etc/cmcluster/3node.ascii -n sysA -n sysB -n sysB
#This insures that your configuration looks good across the nodes. If any problems, fix it now.
7) cd /etc/cmcluster
8) compare the new cluster ascii file with the existing file, and make the relevant changes
9) cmcheckconf -v -C /etc/cmcluster/
10) cmapplyconf -v -C /etc/cmcluster/; cmruncl
11) sysC>cd /etc/cmcluster; mkdir pkgdir
12) rcp/ftp the package control file[s] to sysC
13) cmapplyconf -v -P package_conf_file
14) cmviewcl -v

Might also be worthwhile reading teh Managing MC/ServiceGuard manual available at:
http://docs.hp.com/hpux/ha

[Sanjay_6]

Hi,

You can add a new node to the running cluster. look at the thread below for more info.

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90024/B3936-90024_top.html&con=/hpux/onlinedocs/B3936-90024/00/00/43-con.html&toc=/hpux/onlinedocs/B3936-90024/00/00/43-toc.html&searchterms=add%20node&queryid=20011009-155430

Look for the section "Using MC/ServiceGuard Commands to Add Nodes to the Configuration While the Cluster is Running " for the procedure on adding a new node to the running cluster.

Hope this helps.

thanks

[Sridhar Bhaskarla]

Andrew,

Looks like you have scheduled downtime already for this.

I agree with your procedure.It's clean as you are deleting the cluster and reconfiguring. I would like to add one this before making changes to the ascii file.

cmquerycl -C /etc/cmcluster/newconfig.ascii -n sysA -n sysB -n sysC

instead of editing your own file. Also this gives your some idea about how the systems treat your configuration (For ex., once my systems thought the public LAN as heartbeat and heartbeat LANs as public.Then I replaced the heartbeat switches with faster ones). Edit this file and make modifications.

-Sri

[Santosh Nair_1]

You can add the new node to the cluster without taking either the cluster or any of the packages down. However, you have to make sure that the cluster lock disk and the VGs are shared between the three nodes, although you could probably do without the cluster lock disk.

-Santosh

[Andrew R.]

Thank you all for your reply,much appreciated,
When I ran

sysA>cmquerycl -C -c 'cluster-name'cmclconf.ascii -n node sysA -n sysB -n sysC
I got an error
cmquerycl:Node sysC does not have cluster lock volume group /dev/vgsysA-B
and it didn't generate the new template !!
I think because sysC doesn't share the same disk on sysA I will have to halt the cluster and change the lock disk as I have different lock disk that I can use that will be shared between the 3 nodes but it is different than the one in existing configuration.
Would it work if I just do create the ascii file with hand and without having lock disk for the third node and do
cmapplyconf -C ###.ascii (It might cause problems in future !!)

thanks.

[Carsten Krege]

You need to make sure that sysC has physical access to the cluster lock disk and that you imported the cluster lock VG on sysC. If a cluster lock disk is configured, all nodes *must* have access to it.

You can also run a three node cluster without having a cluster lock disk, too. To do this, you would need to do the following

Halt the cluster
# cmhaltcl

Delete the current config
# cmdeleteconf

Query a new config
# cmquerycl .... (your command)

After editing the cluster.ascii
# cmapplyconf -C cluster.ascii -P package1.conf -P ...

Carsten