Hi all,
I want to upgrade a clustered AIX 5.3 TL10 with PowerHA 5.5.0.4.
To both AIX 6.1 and PowerHA 6.1 in an wasy way.
Any guidelines or procedures in an easy way to accomplish this task safely?
Note : fdown time is an acceptable option.
You sure have searched the forum and the internet for answers to your question because you know that "Google is your friend", no? You sure have found this thread about exactly your question and you surely have some convincing reason why this can't be applied to your case and you need some different solution, haven't you?
Care to tell us this reason?
I hope this helps.
bakunin
The problems you face are less with AIX and PowerHA - more with verifying that your application will run on AIX 6.1.
Bakunin covers when it is known that the application will work on an upgraded version of AIX.
My basic steps are - without including of Bakunin's method for keeping the HACMP/PowerHA monitored applications up as long as possible -
Before doing anything else: backup rootvg, application vg(s), and make a snapshot of PowerHA/HACMP setup; then:
a) update the application - verify it continues to work as expected
b) update AIX
c) update PowerHA
b) and c) sometimes need to be reversed - for example, when the current level of HACMP is not supported on the new version of AIX. Then I would try to update "the current release" of HACMP to latest patch levels so that a newer version AIX is also supported, and then proceed to AIX update, and do a major update of HACMP/PowerHA after the AIX update.
After b) verify the application is still operating as expected (HACMP/PowerHA is not yet changed, so any issues should be AIX level related - updating using alt_disk_install or multibos can help speed reboot and verify issue is OS level related).
After applications are successfully migrated to a new AIX level would I start with the actual HACMP/PowerHA updates.
Ideally, you will have a test node to test application updates and AIX update combinations rather than on production clusters. As these test should have nothing to do with HACMP/PowerHA it should be reliable to test them "outside" of a cluster. When modifications to start/stop/restart scripts are needed, than some testing in a cluster environment is needed.
As is taught in the HA classes: success with any availability assurance product is a 5 step plan:
1) plan
2) plan
3) plan
4) plan
5) implement
Hope this helps!