Re: System backup in a 24/7 environment. -- MIDRANGE-L

Kenneth,

In a message dated 98-06-25 14:31:08 EDT, you write:

> In a 24/7 environment when do you apply delayed PTF's?
>  When and how often do you run SAVSYS?
>  Are you using Save While Active (or should it be called "save SOME while
>  active!)?
>  How do you do release upgrades?

I'd be interested in this as well!  We're currently in a 24/6.5 environment,
and have found no way to push it to 7.  The system must be down for PTF's in
order for the microcode being run to be refreshed.  Same for release upgrades.
Our Systems Engineering group is afraid to even _try_ a RCLSTG for fear that
even a 3-day weekend won't be long enough -- and we're getting damaged objects
because of it.

Due to the volume of data on our 530, "save while active" turned into "crash
3rd shift while saving" because the synch point process kept the objects
locked for too long.  Pushing the wait time up on objects was not an option,
because the folks on the plant floor started turning off their devices and
turning them back on -- causing _WORSE_ problems.  We ended up turning off
mirroring to another box, performing backup on the mirrored box, and then
turning mirroring back on.

While these problems with necessary administrative tasks are irritating at a
manufacturing facility, they can usually be performed on the next 3-day
weekend.  Unfortunately, there _are_ no three day weekends at hospitals and
emergency services organizations -- _BIG_ users of AS/400's as well.  When I
used to do a lot of hospital work, it was no big deal at first to take down
the /400 for a weekend.  The patients didn't _mind_ getting a bill two days
late ;-)!  As time progressed though, we were tied first into the Pharmacy,
then the lab, then the insurance companies, then a smaller hospital that
couldn't afford its' own computer -- as you can well imagine, three-day
weekends were soon not always an option.

I remember hoping that, with some mirroring enhancements made as part of the
V2R2 OS/400 release, that we would be able to shut down mirroring, install an
OS/400 upgrade to the mirrored side of the system, then restart mirroring and
apply the upgrade during the next IPL (an easier window to achieve).  No dice.

This is an operability area that IBM _REALLY_ needs to address, IMO.  With
today's multi-processor systems, there _OUGHT_ to be a way (realizing a
possible failure window exists, which could be addressed combining mirroring
with high-availability in shops that need 24/7) to install OS/400 or PTF's on
a mirrored system by "hi-jacking" one or more of the processors, shutting down
mirroring, loading the upgrade on the mirrored side, refreshing the microcode
on the "hi-jacked" processors and start shunting all new work to them,
refreshing the production processor microcode as they become inactive and
allow new jobs to start on them after they've been refreshed, then performing
a "reverse" mirroring of new OS/400 objects to the production drives when
mirroring is restarted.  Any required object conversion could occur as the
objects became available if it happened quickly enough.  System performance
would degrade exponentially during the process, but at least there would be a
system available to the production users.

Easy?  Probably not.  Possible?  Yes, although basic architecture changes may
be required.  Probable?  Only if enough users clamor for the ability and base
their buying decisions upon it...

JMHO,

Dean Asmussen
Enterprise Systems Consulting, Inc.
Fuquay-Varina, NC  USA
E-Mail:  DAsmussen@aol.com

"The game of life is not so much in holding a good hand as playing a poor hand
well." -- H.T. Leslie
+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---