RM/Cobol Error code 98's

Hi,

I am new to this board, but really am hoping that you will be able to assist
me.

Bottom line question, what in your opinion causes an error 98 to occur using
RM/Cobol ?

For anyone not familiar with rm/cobol error numbers a 98 is Invalid
File/Index Structure and is rectified by running a simple recory procedure
that rebuilds the index blocks on the file.

A little background to this, we have a customer who is having persistent
error 98's on primarily one file on the system.  It is a very active file -
a potential 100 writes per second on occasions, but generally I'd guess
around 20 writes per second would be more realistic.  175 users could be
hitting the file at any time.

It's not a greatly complex file, one main key, one alternate key a varying
record lengths up to 256 bytes.

The customer could work for a w with no errors, but then could also get
errors two or three times a day.
Cobol objects are not changed on the system at all.

Without going into too much detail, system is running on an IBM p650, AIX
Version 5.2, RM Cobol Version 8 (atomic IO is NOT enabled).   Disk array for
data sotrage is connected via fibre optic and IBM switch
Incidently, we have chosen not to enable atomic IO becuase this essentially
hides the problem rather than cure it.

I know that I am probably asking a lot, but any gut reaction / experienced
knowledge would be greatly appreciated.

As you may suspect, the customer is adamant that it is cobol software
causing the problem, we are adament that it's their hardware... we are
rapidly reaching a stalemate which will be no benefit for anyone :-)

I come from a technical background, so any questions/clarifications you
require, please fire away !

Best Regards

Keith Lowe

"Keith Lowe" <keith@support.roadrunner.uk.com> wrote in message
news:42402c2e.0@news.roadtech...
[snip]
> Bottom line question, what in your opinion causes an error 98 to occur
> using
> RM/Cobol ?
[snip]
> I know that I am probably asking a lot, but any gut reaction / experienced
> knowledge would be greatly appreciated.
>
> As you may suspect, the customer is adamant that it is cobol software
> causing the problem, we are adament that it's their hardware... we are
> rapidly reaching a stalemate which will be no benefit for anyone :-)
[snip]

Keith,

I am answering this here for the benefit of other readers, and the archive,
but will handle the actual problem through our support organization, where
your problem is already in our database.

From your description here, I would tend to think this is somehow related to
the hardware/operating system/OS configuration, rather than anything you are
doing in the RM/COBOL program, or anything we (i.e. Liant) are doing either.
I can certainly attest to literally thousands of users on AIX boxes running
mission critical software without this type of problem.  I also know that
certain networking settings on another vendor's operating system (which
shall remain unnamed, but it is *very* popular on the desktop) can produce
similar symptoms.

It is certainly the COBOL software that is *reporting* the problem, and your
customer may not be being too much more analytical than that.  With that
thought in mind, perhaps it might be reasonable to enable atomic I/O so that
the problem can be self-healing, and you get it out of the customer's eye.
I think we can enable logging in such a way that the error will still be
noted, allowing you and us to work on resolving the actual problem with less
impact on your customer's work (ENABLE-LOGGING=ATOMIC-IO).

Be assured that Jack and his staff in London, as well as the entire support
staff here in Austin, will help you get to the bottom of the problem.

Best regards,
Tom Morrison
Manager, Technical Support and Services
Liant Software Corporation