for applying the patches in cluster nodes.

[gb karki]

Hi Expert,

We have 2 nodes SG cluster and SAP/oracle is running after installation we are first time going to apply patches (March-09).

What are the steps i need to follow up and need to take precaution.

Regards
Karki

[Suraj K Sankari]

HI,
Before applying patch take full Ignite backup of both the boxes. after taking the Ignite make very much sure about the backup or if possible take one more copy of the same.
Then move all the packages into Node B halt the cluster into Node A then apply patch into Node A after patching reboot Node A, check everything is OK then start your cluster and switch back your package into Node A check the status for 1 or 2hr. Then go for second Node. Move all package into Node A halt the cluster into node B then apply the patches..into Node B restart the server and start your cluster and move your packages into Node B.

suraj

[Bart Paulusse]

Create an ignite recovery archive (on bootable tape or on your ignite server)
Make an offline backup of your database (or online + all offline redolog files created during backup)
Make a filesystem backup of both your nodes (not including the sapdata files)

Stop serviceguard, place debug files in serviceguard directories on both nodes.
Apply patchbunlde on node 1 and reboot the node.
Check if you can start serviceguard on that node (debug file makes sure oracle and sap do not get started) and check for other errors.
Now apply patches on node 2 and reboot. Again check for errors.
Start SG (still in debug mode) and check for errors (in cluster logging)
Perform a few (depending on available time) of your switch tests. At least check if SG still fails over if you halt node 1, and back...
If all is satisfactory, stop SG, remove debugfiles and start SG.
Check if Oracle and SAP come up normally and check alertlog and sap syslog.

PS. Make sure you have a clear and tested recoveryplan available.

Regards,
Bart

[likid0]

Hy,

I can't find Duncans great post on recomendations when patching, so I'm just pasting:


- Ensure you've done full dependency analysis on all the patches to be installed. The ITRC patch systems or swa should do all that for you now.

- Review the "special installation instructions" for all patches that you are to install.

- Ensure there are no unconfigured filesets on the system... something like this should return nothing:

swlist -l fileset -a state | sed -e '/^#.*$/d' -e '/^$/d' | grep -v configured

- ensure that you at least review a run of "swverify \*" - in general you don't need to worry too much about the WARNING levels, but you shoulkd always try and resolve the ERROR levels.

- if you have non-HP kernel filesets (such as EMC PowerPath or IBM atdd tape drivers), you should always review the patches you are installing, and the patch process with the relevant vendor.

- make sure you have run a "cleanup -d" on the dpot, so you don't have any superseded patches in it.

- always run a preview of the swinstall process and resolve *all* reported issues before proceeding

With those precautions in place you should really never encounter an unbootbale kernel as a result of patching.

[gb karki]

Hi Suraj,

I have moved the packages from node A to Node B.

From Node A:- cmhaltpkg
From NOde B: cmrunpkg

But how to halt the cluster into node A.

Regards
Karki

[Bart Paulusse · Accepted solution]

Hi Karki,

cmhaltnode -v