v5r4m0 to v5r4m5 upgrade speed

We upgraded one of our two systems from v5r4m0 to v5r4m5 this weekend. All went well. The only complaint is that it took longer than I'd hoped! (:

We need to upgrade our 2nd machine in a month or so. I'm looking for ideas that can speed up the process.

We ran the "pre-upgrade verification tool for IBM i." These fixes need to be performed prior to the Big Weekend!

All CDs and PTFs are pre-loaded in Image Catalogs.


THE PROCESS FROM THE TEST MACHINE

Kick all the users off and go to Restricted State

*1 SAVLIB QGPL
*2 SAVLIB QUSRSYS Supposedly 'our' stuff isn't touched, but?

*3 RCLSTG *DBXREF Was only required for the test system
Only took about 10 minutes!

*4 Apply PTFs *PERM One extra IPL here!

Change System Values for QUSRLIBL and QSYSLIBL &c

Manually install Licensed Internal Code took about 1/2 hour

Install OS/400 took about 10 minutes

*5 Sign on to system Sat on C900-2A95 for over an hour!
C900 2A95 [After] Abnormal [Termination] Work Control Block Table cleanup
We have 19 job tables on our test machine, 15 on production
Apply new PTFs (cume) Only for 5722999 and 5722SS1 - the instructions say so!

Apparently, after a 'fresh' install, the system wakes up on the "A" side; the PTFs apply to the "B" side immediately. This means you can't load the Group PTFs as well and apply them in a single IPL...

IPL to the "B" side

Apply Group PTFs Only for 5722999 and 5722SS1!

IPL to the "B" side

Apply v5r4m5 'stray' PTFs The ones that replace v5r4m0 PTFs

Change system values back to 'our' values

*6 IPL to the "B" side This starts everything up with our controlling subsystem and our library lists


We still need to apply the full cume and full Group PTFs to the system, and verify that we have a full set of the v5r4m5 'stray' PTFs


TIME SAVERS

*1
*2
*3
*4 These 4 items are 'no brainer' do-ahead-of-time items.

*3 We have some duplicate items on our production machine to deal with ahead of time.

*5 C900 2A95 [After] Abnormal [Termination] Work Control Block Table cleanup. We have 19 job tables on our test machine, 15 on production

We can compress the job tables, but I think it's more an issue of cleaning up all those left-over spooled files so the jobs go away and there's not so much to clean up. If we minimize the jobs on the system I think things may go faster.

I know we could tell the system to 'detach' the spooled files, but that means finding them will be a lot harder. We currently have 84000 jobs on our production machine. Many of these are job logs. We only have 34000 jobs on our test machine.

We do purge job logs, but keep a rolling week on line for the programmers? We WILL purge a lot more before the Big Day!

*6 We should be able to stack the Group and v5r4m5 PTFs in a single apply-- saving another IPL.

THOUGHTS

I think the process on The Big Day can be shrunk by at least 2 IPLs, and the long delay for the [After] Abnormal [Termination] Work Control Block Table cleanup can be reduced by removing a lot of jobs from the system. These items alone may save close to 3 hours?

My thoughts now are to re-apply the Cume and the Group and the v5r4m5 PTFs to the machine we just upgraded to pick up any 'strays' we missed.

I also want to apply the Cume and the Group PTFs to the 2nd (production) machine ahead of time, even though we will need to re-apply the Cume and the Group for 5722999 and 5722SS1 after the upgrade. This will eliminate the need for an after-upgrade round of PTFs (except the v5r4m5 batch).


YOUR THOUGHTS

If anyone has any ideas of things that will make the process go faster the next time around, please let me know!




Paul E Musselman
IT Technical Support
General Cable Corporation
(859) 572-8030
(859) 572-8477 (fax)
PMusselman@xxxxxxxxxxxxxxxx