[Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

[Martin Bowes]

We have a culprit!

We did it!

We have an OME function which created an optimizedb call. It relied on doing a system(shellScript.sh). The trouble was that this database was moved to a new installation and the shellScript.sh did not make the trip.

As a consequence it then returned 'sh: .../shellScript.sh: No such file or directory' and somehow managed to write that information over the data file.

I am moving that script into position on the new installation as well as adding a stat() call to the OME so that it checks to see if the script exists and is executable.

Marty

PS. I'm actually a *LOT* happier about finding this out. A mystery disk corruption really gnaws at your confidence.

From: Paul White [mailto:paul.white at shift7solutions.com.au]
Sent: 10 October 2018 08:49
To: info-ingres at lists.planetingres.org
Subject: Re: [Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

Hi Marty,

Is key compression stable?
About a year ago I worked on a problem in 10.2 where a simple transaction rollback left the DB inconsistent.
The work around was to avoid using key compression on an index.

Original issue: a normal operational transaction roll back failed.
E_DM0055_NONEXT
E_DM968E_DMVE_BTREE_FIND
E_DM968F_DMVE_BID_CHECK
E_DM964E_DMVE_BTREE_PUT
E_DM967A_DMVE_UNDO
E_DM9639_DMVE_UNDO
E_DM9509_DMXE_PASS_ABORT
E_DM0055_NONEXT
E_DM968E_DMVE_BTREE_FIND
E_DM968F_DMVE_BID_CHECK
E_DM964E_DMVE_BTREE_PUT
E_DM967A_DMVE_UNDO
E_DM9639_DMVE_UNDO
E_DM943A_APPLY_UNDO
E_DM943E_RCP_DBUNDO_ERROR
E_DM943B_RCP_DBINCONSISTENT
E_CL0F0F_LG_DB_INCONSISTENT

Then cycling
E_DM900C_BAD_LOG_BEGIN
E_DM9500_DMXE_BEGIN
E_CL0F0F_LG_DB_INCONSISTENT
E_DM900C_BAD_LOG_BEGIN
E_DM9500_DMXE_BEGIN


So I tried Rollforwarddb but failed with
E_DM943E_RCP_DBUNDO_ERROR
E_DM943B_RCP_DBINCONSISTENT
E_CL0F0F_LG_DB_INCONSISTENT
E_DM0152_DB_INCONSIST_DETAIL
E_DM9015_BAD_LOG_WRITE
E_DM9215_DM0L_PUT   Error
E_DM0100_DB_INCONSISTENT
and
E_DM9352_BAD_DMP_NOT_FOUND
E_DM1307_RFP_RESTORE


Our Ingres support guru had a magic parameter for rollforwardb which allowed us to restore all but inconsistent index. Dropped and recreated the index to fix the problem.
So, we don't use key compression.

Paul



From: info-ingres-bounces at lists.planetingres.org [mailto:info-ingres-bounces at lists.planetingres.org] On Behalf Of Martin Bowes
Sent: Wednesday, 10 October 2018 5:12 PM
To: Laframboise, André (BAC/LAC); info-ingres at lists.planetingres.org
Subject: Re: [Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

We tried.

It didn't.

I was disappointed.

Marty

From: Laframboise, André (BAC/LAC) [mailto:andre.laframboise at canada.ca]
Sent: 09 October 2018 13:53
To: Martin Bowes; info-ingres at lists.planetingres.org
Subject: Re: [Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

Hi,

Errors look related to an indexing structure problem. Data pages themselves seem to be intact. A verifydb -table operation may have also fixed it.

André Laframboise

Conseiller Principal Base de Données, Direction générale de l'Innovation et du Dirigeant principal de l'information
Bibliothèque et Archives Canada<http://www.bac-lac.gc.ca/fra/Pages/bac-web.aspx> / Gouvernement du Canada
andre.laframboise at canada.ca<mailto:andre.laframboise at canada.ca> / Tél. : 613-298-1346

Senior Database Advisor, Innovation and Chief Information Officer Branch
Library and Archives Canada<http://www.bac-lac.gc.ca/eng/Pages/lac-web.aspx> / Government of Canada
andre.laframboise at canada.ca<mailto:andre.laframboise at canada.ca> / Tel: 613-298-1346

[Image removed by sender. http://www.bac-lac.gc.ca/eng/events/PublishingImages/banner.jpg]<http://www.bac-lac.gc.ca/eng/events/Pages/events.aspx>

From: info-ingres-bounces at lists.planetingres.org <info-ingres-bounces at lists.planetingres.org> On Behalf Of Martin Bowes
Sent: Monday, October 8, 2018 6:46 AM
To: Martin Bowes <martin.bowes at ndph.ox.ac.uk>; info-ingres at lists.planetingres.org
Subject: Re: [Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

And an addendum...

I copied the data to disk and loaded it into a temp database. The expected number of rows  came with it and the verifydb said the same thing on the temp database.

I modified the table to reconstruct with nocompression and the verifydb problem disappeared.

So I've decided that the message is probably more incendiary than real.

Marty

From: Martin Bowes [mailto:martin.bowes at ndph.ox.ac.uk]
Sent: 08 October 2018 10:51
To: info-ingres at lists.planetingres.org<mailto:info-ingres at lists.planetingres.org>
Subject: [Info-ingres] W_DM500C_ORPHAN_DATA_PAGE

Hi All,

Recently had a need to run verifydb over every table in a  database. How concerned should I be about:

W_DM502B_INVALID_BT_KIDS   WARNING:  index/leaf page 1 is unreadable because bt_kids has invalid value
of 511.  (Valid values range from 0 to 510.)
...
W_DM500C_ORPHAN_DATA_PAGE  WARNING:  data page 326618 is an unreferenced page (or orphan).  All data on this
page or any overflow pages chained to it is effectively lost.

How concerned should I be given these are warnings?

The table in question is a btree unique. It has data and key compression.

Martin Bowes

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.planetingres.org/pipermail/info-ingres/attachments/20181011/3d2ce9ee/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 752 bytes
Desc: image001.jpg
URL: <https://lists.planetingres.org/pipermail/info-ingres/attachments/20181011/3d2ce9ee/attachment.jpg>