ADSM-L

Re: Core dump / crash

2015-10-04 17:41:28
Subject: Re: Core dump / crash
From: Thomas A. La Porte [mailto:tlaporte AT ANIM.DREAMWORKS DOT COM]
To: ADSM-L AT VM.MARIST DOT EDU
Richard had made a good point late last week regarding keeping up
with the maintenance releases, to which I had wanted to add a
small caveat, but I couldn't think of how to word my own
response.

On Fri, 16 Jul 1999, Richard Sims wrote:

>Folks: If you are running old versions of the client or server software,
>before reporting problems you really need to advance to the latest levels.
>Many of the people who write in are needlessly dealing with problems that
>were resolved some time ago.  Don't waste your time - upgrade your
software.
>At least when you are at the latest levels you have a solid basis for
>seeking IBM resolution of any problems you encounter with current levels.
>
>    Richard Sims, BU

And then late this afternoon, along came the ideal response to
make my point:

On Mon, 19 Jul 1999, Tom Brooks wrote:

[Expiration processing bug info deleted.]

>IF YOU HAVE ALREADY APPLIED 3.1.2.30 and you are exposed
>
>1) Halt the server
>2) Set the server option EXPINTERVAL 0 (in the server options file).
>3) Re-Start the server
>
>We are working to correct this as soon as possible
>
>Tom Brooks
>ADSM Server APAR Coordinator
>ADSM Server Development

The quality control of the ADSM Server development has been a
topic of discussion of this listserv on more than one occasion in
the past several months. Due to the numerous issues that the
level 3 server have had, ranging from the reclamation processing
problem, to this new expiration problem, we have become more and
more reluctant to upgrade to the latest levels of code. Although
these new levels of code often address known issues, it seems
that just as often they introduce new problems of their own,
either in the form of bugs such as the one described, or as
"features," such as the security problem that Martha mentioned in
her message earlier this morning.

It is disturbing to me that these types of very severe problems
in the server code have found their way out the door and into
general release, and it boggles my mind when I read of customers
downloading *fixtests*--which are basically candidates for
release, designed to address a particular customer's problem,
that have gone through even *less* testing than the code that is
actually released--for use on production servers.

We have now implemented a testbed RS/6000 to serve as a platform
to test the new ADSM server software. We can no longer afford to
implement untested software on our production machines, as
relying on the IBM testing process is no longer a dependable
method.

Although there may be many known problems that are fixed in these
newer releases, we'd prefer to stick to the devil we know, rather
than gamble on the unknown problems of the absolute latest
release.

So, while upgrading may often provide a solution to a particular
problem, do keep caution in mind when upgrading your production
platforms to software that no longer appears to be a rigorously
tested product.

 -- Tom

Thomas A. La Porte
DreamWorks Feature Animation
tlaporte AT anim.dreamworks DOT com
<Prev in Thread] Current Thread [Next in Thread>