Neil Truby wrote

> Of course. The client has an EMC SAN which we maintain for
> them, and do all
> sorts of sexy test copies and fantastic "instant"
> point-in-time recovery,
> all of which you can read about in the White Papers section
> of the fantastic
> webiste www.ardenta.com ... but I digress ...
>
> That is an excellent thought, but unfortunately it would only apply if
> problems were encountered with 9.40 straightaway. Suppose we
> converted on
> the Saturday then on the Wednesday found problems in a
> critical business
> process that required a back-out .... we'd be stuffed.
>
I feel the customer is placing too much emphasis on failure. If sufficient testing has been done, the downed system support which is
second to nine should kick in and be sufficient in case of distaster.

Easy to say sitting here I know.

Second option, keep different versions of the data available under both environments using 2 instances and masses of replication.
( this is how we are doing the move from 7.3 to 9.4 next month, if all goes well) But as that is a bit easy, THEY have decided to
change all applications and OS and everything they could think of at the same time.

Colin Bull
c.bull@videonetworks.com

sending to informix-list

"Colin Bull" <c.bull@videonetworks.com> wrote in message
news:btmgbk$3le$1@terabinaries.xmission.com...
> > That is an excellent thought, but unfortunately it would only apply if
> > problems were encountered with 9.40 straightaway. Suppose we
> > converted on
> > the Saturday then on the Wednesday found problems in a
> > critical business
> > process that required a back-out .... we'd be stuffed.
> >
> I feel the customer is placing too much emphasis on failure.

This customer is utterly dependent upon its Informix databases for its very
viability. It cannot survive for long without them. They are *very*
cautious about testing, and about regression plans. And, given the number
of bugs that transpired already to be known to IBM that they ran into in
successive testing on 9.30, 9.40 FC1 and 9.40FC2, this caution has been
vindicated.

>> If sufficient testing has been done, the downed system support which is
second to nine should kick in and be sufficient in case of distaster.

I don't follow this. All the testing in the world doesn't mean you might
not have to regress due to an unforseen circumstance. Unless you mean
testing of the regression path. Which is back where we started: we have
tested it, and it doesn't work.

I agree with you about the high quality of the down system team, although in
the past it has sometimes been hard to get them to act out-of -hours (" ...
We have no record of you on our systems. Which state is that in? ... ").
But, no matter how good they are you'd want to be confident that your system
is actually susceptible to a fix by the engineers, and indeed you'd have
wanted to test it with them. Perhaps this was your point?

> Second option, keep different versions of the data available under both
environments using 2 instances and masses of replication.
> ( this is how we are doing the move from 7.3 to 9.4 next month, if all
goes well) But as that is a bit easy, THEY have decided to
> change all applications and OS and everything they could think of at the
same time.

Offhand I can't think of any reason why this might not work theoretically.
But it would take an enormous amount of resource in terms of extra disk, a
spare server perhaps and, most of all, yet more testing. There's a good
argument to wait until the reversion functionality works as advertised I
think.

Thnaks for your interest, there's some valuable food for thought there.
Keep it coming!

cheers
Neil