The techniue I used on my most recent upgrade was to go from a Cloverleaf 5.2 server (AIX 5.2) to a Cloverleaf 5.6 server (AIX 5.3), which went as well as I could imagine, considering the magnitude of the upgrade (hundreds of sites and over a thousand interfaces).
Here is a high level description of my steps:
– wipe server LPAR clean and do scratch install of everything (OS, HACMP, SNA, Cloverleaf 5.6, make advancements like ether channel)
– do vulnerability scans and comply with any security department findings
– do scratch install of Cloverlead 5.6 and create mda_upgrade site as default site
– NFS mount READ ONLY /sites filesystem from the Cloverleaf 5.2 server onto the Cloverleaf 5.6 server and make it look like it is where Cloverleaf expects the old sites to be via a symbolic link
– use hcirootcopy on the Cloverleaf 5.6 server to upgrade each old 5.2 site and activate them granually to control the size of my potentail problems
– use our downtime model to turn on and test all the modules in the site I upgrade
– make the site ready for golive with alomost 100% confidence and know I can back out quickly with 100% confidence; so I’m not faced with a no point of return situation
– when ready to golive and switch over stop Cloverleaf 5.2 inbound interfaces let outbound interfaces empty out then shutdwon 5.2 site and turn on interfaces on new Cloverleaf 5.6 server
The average downtime of each site golive switch over from Cloverleaf 5.2 server to Cloverleaf 5.6 server ran about 2 mintues most of the time and out of some 1,200 interfaces I was able to do the entire upgrade without virtually a single significant issue that wasn’t detected and remediated during the mock golive testing using our downtime model.
Russ Ross
RussRoss318@gmail.com