Kris Buelens <[EMAIL PROTECTED]> wrote :-
> I work like Jim: never PUT2PROD.
> Copy the runtime minidisks from the install user (eg 5VMTCP30) to
> alternate addresses of the "active" user (eg TCPMAINT).  When the time
> is right, the mdisk addresses are swapped.  This process is first
> tested on the SW Installation system and then repeated on production
> systems. The minidisks passwords tell which is which.
> This way we always have a backout and operators can be explained which
> minidisk addresses to swap and which servers to restart in case a
> backout is required.

As Jim said - there are many different ways of doing maintenance. However, 
the guiding principal is always to have a thought out process where you 
can back out easily.

We have adopted a slightly different variation. We have a 2nd level 6-pack 
"build" system where we do all maintenance and build of base VM and 
related products (only those supplied with VM). As such we can go through 
the complete process, including PUT2PROD. 
We then map these minidisks on our other systems and, this is the 
important point, copy them to release dependant minidisks on the 
production systems. Note, we do not use the original build disks so that a 
rebuild will not affect other systems. 

There are quite a few other processes in our maintenance strategy (like 
holding CMS disks (190, 19D, 19E etc.) on a release dependant userids 
called $$CMSnn, Also rebuilding all segments with CMS release dependant 
names to make staged rollout of a new CMS level possible because both 
segment levels can coexist). There are quite a few other things as well 
but all designed to allow new levels to be phased in with an easy backout 
path (mostly by doing a minidisk switch). The other thing we do is Rename 
out CP modules as CPnnn to indicate its level so that we can IPL a new 
level but easily fall back).

The thing is that our processes may not suit you because they are 
dependant on, and integrated with, other things we do. As long as you have 
a planned process to switch in (and out) new maintenance then you are 
fine. 


Colin Allinson
Technical Manager - VM Systems Support
Amadeus Data Processing GmbH

Reply via email to