Re: PTF Best Practice? apply to separate LPAR first, or to a non-production machine?

Even though we have 10 lpars of i we don't quite do it this way. We have
three lpars on what I'll call our "3" machine. They get new PTF's on a
Monday or Tuesday. We have four lpars of i on what I'll call our "2"
machine. They get the PTF's on the next day or two. The "1" machine has
three lpars of i. It gets the PTF's that weekend. No real exhaustive
test of the PTFs are done.
OS upgrades are different. One of the lpars on the "3" machine is a
sandbox. It gets the upgrade first and some more testing is done. Not
exhaustive by any means. But that OS upgrade may stay on there for a few
months before the others are upgraded.

Remember, IBM does some pretty good testing in their lab. There's always
the possibility that even with exhaustive tests you may run into issues.
There's a reason why IBM does not put ALL PTFs on a cume. Only PTF's
which meet certain criteria make it. A PTF which was put out there for a
few unique people may not make it because of the possibility of unintended
side effects. Perhaps you have this, this and this LPP but not this LPP.
And you have this, this and this hardware. And a particular load, like
receiving a fax while transmitting EDI while generating a PDF file with
transformation services. The end result may be you trigger something that
the lab didn't catch. So, with that thought in mind, your
test/development may not catch something that your live machine catches.
Doesn't preclude a good faith effort to try though.



Rob Berendt