Conference star::wizards

1767.0. "Open: why hanging?" by STAR::JKEENAN () Fri Apr 25 1997 13:46

Return-Path: "VMS001::WWW"@vms001.das-x.dec.com
Received: by vmsmkt.zko.dec.com (UCX V4.1-12, OpenVMS V6.2 VAX);
	Sat, 19 Apr 1997 07:27:53 -0400
Received: from vms001 by mail11.digital.com (8.7.5/UNX 1.5/1.0/WV)
	id HAA30222; Sat, 19 Apr 1997 07:20:25 -0400 (EDT)
Date: Sat, 19 Apr 1997 06:23:59 -0400
Message-Id: <[email protected]>
From: "VMS001::WWW"@vms001.das-x.dec.com (19-Apr-1997 0624)
To: [email protected], [email protected], [email protected]
Subject: Ask the Wizard: '[email protected]'
X-VMS-To: [email protected]

Remote Host: client8289.globalnet.co.uk
Browser Type: Mozilla/2.0 (compatible; MSIE 3.01; Update B; Windows 95)
Remote Info: <null>
Name: Dan Cassidy
Email Address: [email protected]
CPU Architecture: VAX
Version: v 6.2
Questions: 

Dear Wizards,

We have a VAX 3190 computer running a Steel Mill.

Since we upgraded from VMS 6.1 to 6.2 approx. a year ago we have had system
hangs,
the frequency varies but on average approx. once every 2 weeks.

Recently the frequency of hangs has increased, and currently the system tends to
hang 
when we start up the applications.

We have forced a dump from the console and analysed the dump file many times and
we have 
identified some characteristic details of the problem.....


1. The hang is always a run-loop at LCK$COMP_GGMODE+00017, goes forward 7
instructions then
loops back, in KERNEL mode, in the context of an application program.

2. The problem seems to be that the chain of locks attached to a resource seems
to have one of the locks pointing to a previous lock or the same lock ( it
varies ) in the queue.

3. Hence when VMS ( and also SDA ) try to navigate this chain of locks they get
stuck in a run-loop.

4. The hang problem is nearly always in an application program that uses Rdb (
which is heavy on locks ), 
and it is at a point in the application program where Rdb code is entered.

4. The parent of the parent of the parent of the resource in question is a disk
volume - the 
data disk which holds the Rdb database.

5. Looking at the stack, the problem always occurs on a DEQUEUE operation.


Other details.......

a) It is a twin system ( main and standby ) VAX 3190 128Mb memory - hang problem
is same on both.

b) System uses Rdb.

c) System uses X-windows for user interfaces ( problem seems to only happen when
these apps are running )

d) We have re-installed Rdb, problem still the same.

e) We have suspected LOCKIDTBL growth as the problem but have ruled this out.

Yours faithfully,
Dan Cassidy

   Please contact DIGITAL and Oracle customer support directly.