AIX

Hi All,

As per the IBM upgrade/support matrix

IBM Techdocs Technote: PowerHA for AIX Version Compatibility Matrix

we can't do online upgrade or rolling migration from powerha v7.1.0 to v7.1.3. on AIX61_TL9_SP4, So we are following the steps as below ...

1 ) Bring down the cluster
2 ) CAA_FORCE_ENABLED=1
3 ) /usr/sbin/rmcluster -n cluster_name
4 ) upgrade PowerHA using smitty update_all ( updating PowerHA software from 7.1.0 to 7.1.3 ) to all the nodes.
5 ) after upgrade restart all the nodes .. shutdown -Fr
6 ) Create the CAA cluster by running the /usr/es/sbin/cluster/utilities/clmkcaa command
7 ) Verify that the CAA services are active by running the /usr/sbin/lscluster -m
8 ) If step 6 failed then .. and cluster don't form properly then follow the below steps ..
/usr/es/sbin/cluster/conversion/clconvert_snapshot -v version -s snapshotname ( version is the previous version before upgrade can be taken using halevel command )
E.g. /usr/es/sbin/cluster/conversion/clconvert_snapshot -v 7.1.0 -s snapshotname
From the command line, enter smit hacmp.
In SMIT, select Cluster Nodes and Networks > Manage the cluster > Snapshot Configuration > Restore the Cluster Configuration from a Snapshot, and press Enter.
Select the snapshot that you want to restore, complete the required fields, and press Enter.
9 ) Run the /usr/sbin/lscluster -m command to verify that the Cluster Aware AIX (CAA) services are active on each node in the cluster.

So far this worked well, but few servers we aren't getting downtime for powerha upgrade, is there any one done online or rolling migration ? Any one found any clever way to over this downtime difficulty ?

Thanks In Advance,
Amrit

I have not much with PowerHA/SystemMirror - very first versions such as 7.1.0 to 7.1.1 and later - HOWEVER - I do recall there is a fundamental change in the way CAA (cluster aware AIX) is implemented that made it impossible to update 7.1.0 to 7.1.1, The only path was remove, and start again. I have heard that has been softened a bit - but I prefer clean and tend to not use migration/upgrade paths in general.

A basic feature in PowerHA has been the ability to stop the cluster while leaving the application running. Will this (leaving applications "unmanaged by PowerHA", but operational minimize your downtime sufficiently?

PowerHA/SystemMirror from 7.1.0 to 7.1.3 - what else are you updating? Won't those updates require some maint. time as well? Why not combine them? imho it is not fair to put a requirement for downtime for maint solely on the OS and/or middleware. Applications need maint. as well. The 'business' should know/understand that - and sign off for the impact of unexpected downtime that could have been prevented if maint had been timely. (Or do you wait for your engine to lock up before going to the garage for an oil change?).

Thank you very much MichaelFel, can you please tell ( if you have doc/link handy ) what is the fundamental diff of CAA between 7.1.0 and 7.1.1 ?
And what if .. i do follow the below steps ...
1 ) Put the cluster at UNMANAGE STATE
2 ) CAA_FORCE_ENABLED=1
3 ) /usr/sbin/rmcluster -n cluster_name
4 ) upgrade PowerHA using smitty update_all ( updating PowerHA software from 7.1.0 to 7.1.3 ) to all the nodes.
5 ) Create the CAA cluster by running the /usr/es/sbin/cluster/utilities/clmkcaa command
6 ) Run Sync and Verify the cluster.
7 ) Move all the RGs from one Node1 to Node2, reboot Node1
8 ) Once Node1 rebooted start cluster Services on Node1 and move all the RGs from Node2 to Node1 and reboot Node2
9 ) Once Node2 is rebooted, Start cluster services and redistribute the RGs.
This sounds me bit CRAZY ..