Take a long long long hard look at MFILES...... 

I haven't seen *this* problem in that context but a few possibilities come
to mind. BTW: 9.6.2.(late) wouldn't hurt...

Record locks are (as John said) a distinct possibility - and you may want to
check the behaviour of locks gained outside a transaction boundary and then
re-acquired and updated inside that boundary. If you have just added
transaction statements you may be surprised at what can happen to record
locks that get subsumed/promoted inside a transaction boundary (there were
some changes in this area - can't *quite* remember the release but it's
somewhere around here.

Otherwise:

Use GCI? - burn file handles possibly? (open and no close, but also (maybe)
re-open them again and again etc....

Any triggers?

Opening OS level files and not closing them ?

Insufficient gap between MFILES and MAXFILES ? (bearing in mind any
additional file handles you may use in GCI)

Worth adding an error trap on your "writes/deletes" just in case.....

Let us know what you find

Regards

JayJay

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of John Hester
Sent: 12 November 2004 21:05
To: [EMAIL PROTECTED]
Subject: Re: [U2] DG Intel disk write problems

Tom Dodds wrote:
> We have a DG Intel server running UV 9.6.1 and seem to be missing writes.
> We have a series of writes, 4, in a row within a basic program and
sometimes
> it seems that one of the items is not updated not written.  Has anyone
> experienced a similar problem.  I know it's a long shot, but the customer
> wanted me to ask.

The only time I've seen something like this is when READU locks aren't 
properly implemented in a program.  Could the lock on the item in 
question be prematurely released, say with a WRITEV prior to the final 
WRITE?  This would allow another process to write back a stale copy, 
giving the appearance that the write never happened.

-John
-- 
John Hester
System & Network Administrator
Momentum Group Inc.
(949) 833-8886 x623
http://memosamples.com
-------
u2-users mailing list
[EMAIL PROTECTED]
To unsubscribe please visit http://listserver.u2ug.org/
-------
u2-users mailing list
[EMAIL PROTECTED]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to