At 9:04 AM -0700 3/15/01, Burgess Peter wrote:
>Hi
>
>We are currently investigating some issues that seem to re occur every 6
>months. The database analysis shows a number of broken chains on the image
>sets. I thought that Broken chains in the new order of things would be very
>very rare. We are currently running under 6.0 of mpe/ix.
>
>Does anyone know of a programmatic reason as to how we can cause this. We
>are using Cobol and Powerhouse to talk to the databases. No hardware
>failures on the disks have been reported.
>
>If you require any further information please let me know.
>
>Regards
>
>Peter Burgess
>MicroWarehouse
>www.microwarehouse.co.uk
>
Ken Paul <[log in to unmask]> posted to HP3000-L two messages that could
explain the problems Peter is experiencing. The first one was thirteen
months ago, on February 24, 2000. The second one was ten months ago, on May
26, 2000. Because of their relevance, I'm repeating them today:
---------------------------------
Return-Path: <[log in to unmask]>
X-Sender: [log in to unmask]
Date: Thu, 24 Feb 2000 18:00:55 -0700
To: [log in to unmask]
From: Ken Paul <[log in to unmask]>
Subject: ATTENTION MPE/iX 6.0 Users! (Long)
Even though this message may not go over big with some folks at HP, I felt
that the risk to users was too great to not post a heads-up when there is a
current (pain free) work around to the current problem.
Since the beginning of the year we (Adager) have dealt with at least 2 sites
who have experienced system failures/hangs and upon rebooting there machines
have had problems with their IMAGE databases. The 2 cases that I worked on
personally had the symptoms where one of the datasets had holes/missing
records. Both the users and myself thought that this was rather odd and
should be prevented by the Transaction Manager(XM).
While attending SigSoftVend last week I was talking to a CSY lab engineer
and mentioned these 2 occurrences. He wondered whether both clients were on
6.0. Long story short, he believes that there is a corner case bug within
the new XM code for 6.0 which he is now going to develop a patch. Until
that patch becomes available, he mentioned that users of 6.0 could utilize a
new feature on 6.0 (and only 6.0) within the XM which would prevent the
problem from happening.
In the 6.0 Communicator on page 216 there is an article on Transaction
Manager. On page 219 within the article there is a section entitled
Checkpoint Improvement which describes 2 new commands in Volutil for showing
and setting this new feature.
In checking the last years worth of messages on 3000-l there have been a few
references to this option. Mike Hornsby asked about it in February of 1999
and Steve Macsisak with HP had a good description of it on February 19th
1999. Carl McNamee mentioned it favorably back in November of 1999.
By default this option is disabled when you install 6.0 and if you enable it
for your volume sets then you should not experience the problem mentioned
above and you may even see a performance improvement on Checkpoints within
the XM. The 2 commands are:
showchkptstat <volset-name>
alterchkptstat <volset-name> ENABLE|DISABLE
A key paragraph on page 220 then states:
"Since this solution remembers the dirty page ranges in a data structure
associated with each open file, it can't affect files that were open before
giving this request (ENABLE). However, it will come into effect for the
files that were open (for the first time) after giving the request. So the
best way to enable or disable the Checkpoint Improvement is to reboot the
machine after giving the command."
Some of you may be wondering why I am posting this information and others
may think that I am acting like Chicken Little ("The sky is falling!").
There are 2 main reasons for my post (and one of them is NOT to start a
firestorm of controversy).
1. I feel that the users should know of a potential data corruption problem
(no matter how obscure (it's not a corner case when it happens to ME!)) and
be given the chance to stop it from happening if they have the means to
RIGHT NOW instead of waiting for a patch from HP.
2. I really don't want to fix more databases than I have to and also have
the VERY UNPLEASANT task of telling a user that some of their data is gone
from their IMAGE database.
I do not know of an SR# for this problem nor a patch id (sorry). As I said
at the beginning, this may not make me very popular with some people but
that is not why I'm here.
This post is coming from me personally and NOT my employer (although my
employer usually backs me up when I make such posts :-)!
Thank you for your time,
+---------------+
| |
| r | Ken [log in to unmask]
| e | http://www.adager.com
| g | Ken Paul Tel 208 726-9100
| a | Customer Support Fax 208 726-2822
| d | Adager Corporation
| A | Sun Valley, Idaho 83353-3000 U.S.A.
| |
+---------------+
--------------------------------------------------------------
Return-Path: <[log in to unmask]>
X-Sender: [log in to unmask]
Date: Fri, 26 May 2000 17:58:13 -0600
To: [log in to unmask]
From: Ken Paul <[log in to unmask]>
Subject: Reminder for ALL MPE/iX 6.0 Users
Just a reminder to all of you who are already on MPE/iX 6.0 or are thinking
of upgrading to 6.0.
This doesn't include anyone on 5.5 or 6.5, only MPE/iX 6.0 with no PP, PP1
or PP2.
I sent a message to the list back in late February about a potential problem
with 6.0 and a possible fix available within VOLUTIL. Since that time we
have repaired a lot of IMAGE databases on 6.0 machines which possibly could
have benefitted from this new VOLUTIL feature. As the old saying goes: "an
ounce of prevention...."
Return-Path: <[log in to unmask]>
X-Sender: [log in to unmask]
Date: Wed, 31 May 2000 16:39:16 -0600
To: [log in to unmask],[log in to unmask]
From: Ken Paul <[log in to unmask]>
Subject: Re: Reminder for ALL MPE/iX 6.0 Users
Not only was my previous message to the list "as clear as mud" but it also
raised an issue that I was not aware of regarding ALTERCHKPTSTAT and
possible SA2182's.
First let me try to state more clearly that the ALTERCHKPTSTAT command is
only in VOLUTIL on ALL flavors of MPE/iX 6.0. It does not exist on 5.5
versions and 6.5 versions of MPE/iX.
So, if you are on MPE/iX 6.0 with no Powerpatch, MPE/iX 6.0 with PP1 or
MPE/iX 6.0 with PP2 you may want to ENABLE ALTERCHKPTSTAT within VOLUTIL for
all of your volume sets.
According to at least one user ("Genute, A Thomas" <[log in to unmask]>),
they experienced an SA2182 after enabling a volume set for ALTERCHKPTSTAT
and were told by HPRC to disable ALTERCHKPTSTAT. This is the first I had
heard of any problems caused by enabling ALTERCHKPTSTAT and I apologize to
anyone else who had this unfortunate experience.
Here is the latest information that I was able to receive from HP.
>1. Multiple occurrences of database corruption after a system abort.
>Transactions are not being properly recovered. This problem exists
>in 5.0 or later. Patch MPEKXN3A is general release for 5.5.
>Patch MPEKXN3B is in beta test status for 6.0.
This was as of March 31, 2000. The patch MPEKXN3B is general release as of
May 31, 2000
Patch MPELX46A is in beta to address the SA2182 issue.
What I would now like to recommend is that if you are on 6.0 you should
check to see if you have patch MPEKXN3B on your system and should get patch
MPELX46A before you Enable ALTERCHKPTSTAT on your volume sets.
I'm not sure how prevalent SA2182 is and as Steve Cole commented, they have
several machines with this feature enabled and have not seen any SA2182s. I
just wanted the list to be aware of all the possibilities available to them.
The bottom line is that I would diagnose your IMAGE databases on a periodic
basis to look for any possible problems that might be hiding in them. It is
no different than going to the doctor or dentist periodically to get checked
out. The health of your database is directly related to the health of your
company!
I hope this helps,
+---------------+
| |
| r | Ken [log in to unmask]
| e | http://www.adager.com
| g | Ken Paul Tel 208 726-9100
| a | Customer Support Fax 208 726-2822
| d | Adager Corporation
| A | Sun Valley, Idaho 83353-3000 U.S.A.
| |
+---------------+
|
