To upgrade SG to A.11.18 ( four node Metro cluster)

[HP-UX_Ali]

Hello Experts,

I have scenario for four node in which i have to configure Metro Cluster.

I have already setup DR site with latest SG-A.11.18 and created local cluster.

In Primary site (which is production) Only serviceguard extension for rac is installed and i have to install Service guard A.11.18 for Metro & serviceguard extension for rac to latest version addition to this i have to re-organize the filesystem.

Now the issue is, Its very critical system (Primary site). Its very hard to get downtime for long hours ( To upgrade both nodes to Latest Sg & patches ..etc).
What will be the best way to perform the upgrade with minimul down time.

Your Reply to this issue will be appreciated.

Thanks & Regards
Ali...

[likid0]

I don't know about metrocluster, but as for SG on your primary site, you should do a rolling upgrade like always.

You can find the info in Apendix E of the service guard manual.

To be able to use a rolling upgrade to update to 11.18 , your SG version has to be at least 11.15 or 11.16.

from 11.18 release notes:

Rolling Upgrade
In some cases you can upgrade Serviceguard and HP-UX without bringing down the cluster;
you do this by upgrading one node at a time, while that nodeâ s applications are running on an
alternate node. The process is referred to as a rolling upgrade, and is described in Appendix E
of the Managing Serviceguard manual.
Requirements for Rolling Upgrade
Rolling upgrade to Serviceguard A.11.18 is supported only if you are upgrading from:
â ¢ Serviceguard A.11.15 or greater on HP Integrity systems running HP-UX 11i v2 or 11i v3;
or
â ¢ Serviceguard A.11.16 or greater on HP 9000 systems running HP-UX 11i v2 or 11i v3.

[HP-UX_Ali]

Hello Daniel,

On the primary side there is no Service guard A.11.17 installed, so i cannot perform rolling upgrade.


Regards

[Rita C Workman]

I thought this sounded familiar...

You already have this same question on another post, with answers:

http://h30499.www3.hp.com/t5/Serviceguard/To-upgrade-SG-to-A-11-18-four-node-Metro-cluster/m-p/5269610#M25386

Rita

[HP-UX_Ali]

Hello Rita,

If you read the link carefully, you can understand well. I have close the thread and started in serviceguard department. and also i have asked the question also but no reply.

Just saying node by node is not the solution, practically it wont work some time unless you have experience. if you have any steps or procedure do let me know.

Below is the question i have sked but no reply....

One more thing, how it is possible DR- with 2 nodes & primary with 1 node MC/SG and again how i will add 2nd Node after upgrade & patches in MC/SG.
There is note in Metro-Rac docs online deletion & adding required same amount of nodes on both sites.

Pls confirm..


Thanks...........

[Rita C Workman · Accepted solution]

Ali,

You said downtime is an issue. You mention a primary side with Oracle Rac, but not how many servers. You mention a DR site running at MC/SG 11.18 and that is up and running and that you set it up.
So I took this as saying that you are experienced.

Without knowing how many servers you have involved at the primary side my suggestion was to create a single node cluster to begin.
Since you created the DR cluster, you would know that the command is cmquerycl -C yourcluster.ascii -n
Then you would edit the file to meet your needs.
You run cmcheckconf -k -v -C yourcluster.ascii
or just cmcheckconf -v -C yourcluster.ascii
Ensure no errors/fix accordingly
When all is good:
cmapplyconf -k -v -C yourcluster.ascii
or cmapplyconf -v -C yourcluster.ascii

If downtime permits then try to add another node into your cluster repeating the above steps with the new server/node included.
cmquerycl -C yourcluster -n -n
Edit ...
cmcheckconf ...
cmapplyconf ...

And of course along the way you would begin adding your packages, so just change the "C" to a "P" and follow the instructions that a specific for packages.

By starting with just a single node cluster you divide and conquer the work. And if any errors happen adding one new server/node at a time you know if there are issues it's on the new server, not the one already in there. The same goes for the packages.

Now if you requre anymore help, you can download information from docs.hp.com (it will send you to the new site). Or just get in touch with my business and we can arrange some remote HPUX services and I'll be happy to do it for you, for a fee.

Regards,
Rita

[Rita C Workman]

..oh forgot to mention..

You don't mention if your using lock disk or if you want to go with a Qorum server, so if you're using a Quorum server you would change the command:

cmquerycl -C yourcluster.ascii -q -n -n -n

/rcw

[HP-UX_Ali]

Thanks Rita,

Good sense of humor, Like the way. By the way the primary side with 2 node cluster.
you mean to take one by one and check for cmquerycl for making up the cluster and after down time add the 2nd node of primary side.

Ok i will do it. Thanks a lot for your effort and patience. and also by the way how can i be in touch with you. if any thing goes wrong in production Implementation time and as you said you will do it for me FREE :)

Lets see......

Regrads
Ali...