Hello community,

how should I interpret these messages in the error log ?



Wed Feb 18 10:15:02 2009 E_DM93AF_BAD_PAGE_CNT Wrong number of pages read or written. At page 144000, an attempt was made to read or write 8 pages, but only 0 pages were actually processed.

Wed Feb 18 10:15:02 2009 E_DM9005_BAD_FILE_READ Disk file read error on database:colibri tabletatext pathname:/uvol5/ingres/ingres/data/default/colibri filename:aaaaaapg.t00 page:144000

Wed Feb 18 10:15:02 2009 E_DM9335_DM2F_ENDFILE End of file was reached during a read or write operation.

Wed Feb 18 10:15:02 2009 E_DM920D_BM_BAD_GROUP_FAULTPAGE Error faulting a group of pages.
Wed Feb 18 10:15:02 2009 E_DM9C83_DM0P_CACHEFIX_PAGE An error occurred while fixing a page in the buffer manager.
Wed Feb 18 10:15:02 2009 E_DM9204_BM_FIX_PAGE_ERROR Error fixing a page.
Wed Feb 18 10:15:02 2009 E_DM9261_DM1B_GET Error occurred getting a record.

Wed Feb 18 10:15:25 2009 E_DM9206_BM_BAD_PAGE_NUMBER Page number on page doesn't match its location.




Hardware support says there is *no* disk problem, but is still checking.

Jürgen

Hi Jürgen,
I recommend that you raise a support call with Ingres Corp.
(I presume that you have a support contract.)

Michael

-----Original Message-----
From: info-ingres-bounces (AT) kettleriver...ting (DOT) com [mailto:info-ingres-bounces (AT) kettleriverconsulting (DOT) com] On Behalf Of Jürgen Cappel
Sent: 18 February 2009 12:55
To: info-ingres (AT) kettleriverconsulting (DOT) com
Subject: [Info-Ingres] Disk problem ?

Hello community,

how should I interpret these messages in the error log ?



Wed Feb 18 10:15:02 2009 E_DM93AF_BAD_PAGE_CNT Wrong number of pages read or written. At page 144000, an attempt was made to read or write 8 pages, but only 0 pages were actually processed.

Wed Feb 18 10:15:02 2009 E_DM9005_BAD_FILE_READ Disk file read error on database:colibri tabletatext pathname:/uvol5/ingres/ingres/data/default/colibri filename:aaaaaapg.t00 page:144000

Wed Feb 18 10:15:02 2009 E_DM9335_DM2F_ENDFILE End of file was reached during a read or write operation.

Wed Feb 18 10:15:02 2009 E_DM920D_BM_BAD_GROUP_FAULTPAGE Error faulting a group of pages.
Wed Feb 18 10:15:02 2009 E_DM9C83_DM0P_CACHEFIX_PAGE An error occurred while fixing a page in the buffer manager.
Wed Feb 18 10:15:02 2009 E_DM9204_BM_FIX_PAGE_ERROR Error fixing a page.
Wed Feb 18 10:15:02 2009 E_DM9261_DM1B_GET Error occurred getting a record.

Wed Feb 18 10:15:25 2009 E_DM9206_BM_BAD_PAGE_NUMBER Page number on page doesn't match its location.




Hardware support says there is *no* disk problem, but is still checking.

Jürgen


_______________________________________________
Info-Ingres mailing list
Info-Ingres (AT) kettleriverconsulting (DOT) com
http://www.kettleriverconsulting.com...fo/info-ingres

Hi Jurgen

It may not be the disk itself, I'll bet the table ptatext has blobs...and there have been some strange things like this surface a few times.

Run a verifydb -mreport -u<owner> -sdbname colibri -oxtables ptatext

And capture the output to a file....

Martin Bowes

-----Original Message-----
From: info-ingres-bounces (AT) kettleriver...ting (DOT) com [mailto:info-ingres-bounces (AT) kettleriverconsulting (DOT) com] On Behalf Of Jürgen Cappel
Sent: 18 February 2009 12:55
To: info-ingres (AT) kettleriverconsulting (DOT) com
Subject: [Info-Ingres] Disk problem ?

Hello community,

how should I interpret these messages in the error log ?



Wed Feb 18 10:15:02 2009 E_DM93AF_BAD_PAGE_CNT Wrong number of pages read or written. At page 144000, an attempt was made to read or write 8 pages, but only 0 pages were actually processed.

Wed Feb 18 10:15:02 2009 E_DM9005_BAD_FILE_READ Disk file read error on database:colibri tabletatext pathname:/uvol5/ingres/ingres/data/default/colibri filename:aaaaaapg.t00 page:144000

Wed Feb 18 10:15:02 2009 E_DM9335_DM2F_ENDFILE End of file was reached during a read or write operation.

Wed Feb 18 10:15:02 2009 E_DM920D_BM_BAD_GROUP_FAULTPAGE Error faulting a group of pages.
Wed Feb 18 10:15:02 2009 E_DM9C83_DM0P_CACHEFIX_PAGE An error occurred while fixing a page in the buffer manager.
Wed Feb 18 10:15:02 2009 E_DM9204_BM_FIX_PAGE_ERROR Error fixing a page.
Wed Feb 18 10:15:02 2009 E_DM9261_DM1B_GET Error occurred getting a record.

Wed Feb 18 10:15:25 2009 E_DM9206_BM_BAD_PAGE_NUMBER Page number on page doesn't match its location.




Hardware support says there is *no* disk problem, but is still checking.

Jürgen


_______________________________________________
Info-Ingres mailing list
Info-Ingres (AT) kettleriverconsulting (DOT) com
http://www.kettleriverconsulting.com...fo/info-ingres

On Feb 18, 2009, at 7:55 AM, Jürgen Cappel wrote:

The file is shorter than expected. It's probably not a failing disk
drive, but it might have been caused by a system crash in the middle
of filesystem writes. I agree, open an issue if you have support.

Karl

Hi Jürgen,

I had one of these a couple of years ago. Solaris, Ingres 2.6 sp2. The
btree index was corrupted and duplicate data entries had somehow gotten into
the table. Bug 114324 was mentioned in the issue (122221) but I am not 100%
sure it was related. I think some of the index pointers were pointing to
empty disk past the end of the table.

I located the faulty data by selecting and grouping by the (supposedly)
unique key fields. Eg

Select k1, k2, k3, count(*) cnt from tab
group by k1, k2, k3
having count(*) > 1


Rollforwarddb did not fix it because the table had been corrupt weeks before
being discovered.
Tried using Copydb to reload the table but it failed with duplicate entries.
I didn’t try usermod.
I manually rebuilt the table with a couple of selects to isolate and clean
the duplicated data. Something like this:

create tabdupes as
select k1, k2, k3, count(*) cnt from tab
group by k1, k2, k3
having count(*) > 1

create tabnew1 as select T.* from tab T
join tabdupes D on T.K1 = D.k1 and T.K2 = D.k2 and T.K3 = D.k3

create tabnew2 as select T.* from tab T
join tabdupes D on T.K1 = D.k1 and T.K2 = D.k2 and T.K3 = D.k3
where D.k1 is null

drop tab
create tab ...
insert into tab select * from tabnew1
insert into tab select distinct * from tabnew2

I highly recommend get Ingres support to check into it to avoid potential
data loss.

Paul

-----Original Message-----
From: info-ingres-bounces (AT) kettleriver...ting (DOT) com
[mailto:info-ingres-bounces (AT) kettleriverconsulting (DOT) com] On Behalf Of Jürgen
Cappel
Sent: Wednesday, 18 February 2009 10:55 PM
To: info-ingres (AT) kettleriverconsulting (DOT) com
Subject: [Info-Ingres] Disk problem ?

Hello community,

how should I interpret these messages in the error log ?



Wed Feb 18 10:15:02 2009 E_DM93AF_BAD_PAGE_CNT Wrong number of pages read
or written. At page 144000, an attempt was made to read or write 8 pages,
but only 0 pages were actually processed.

Wed Feb 18 10:15:02 2009 E_DM9005_BAD_FILE_READ Disk file read error on
database:colibri tabletatext
pathname:/uvol5/ingres/ingres/data/default/colibri filename:aaaaaapg.t00
page:144000

Wed Feb 18 10:15:02 2009 E_DM9335_DM2F_ENDFILE End of file was reached
during a read or write operation.

Wed Feb 18 10:15:02 2009 E_DM920D_BM_BAD_GROUP_FAULTPAGE Error
faulting a group of pages.
Wed Feb 18 10:15:02 2009 E_DM9C83_DM0P_CACHEFIX_PAGE An error occurred
while fixing a page in the buffer manager.
Wed Feb 18 10:15:02 2009 E_DM9204_BM_FIX_PAGE_ERROR Error fixing a page.
Wed Feb 18 10:15:02 2009 E_DM9261_DM1B_GET Error occurred getting a
record.

Wed Feb 18 10:15:25 2009 E_DM9206_BM_BAD_PAGE_NUMBER Page number on page
doesn't match its location.




Hardware support says there is *no* disk problem, but is still checking.

Jürgen


_______________________________________________
Info-Ingres mailing list
Info-Ingres (AT) kettleriverconsulting (DOT) com
http://www.kettleriverconsulting.com...fo/info-ingres

You should check to make sure the file is OK at the OS level.

Shutdown Ingres and try to make a copy the datafile, then see if the copy is the same size as the original.
If there is corruption in the file, you may get an I/O error.

P.S. Don't create the copy in the original directory, copy it to something like /tmp.

If the file is OK OS wise, then try to read TIDs in the pages higher then the one it failed at.
Maybe try a copy out too.

Andre

-----Original Message-----
From: info-ingres-bounces (AT) kettleriver...ting (DOT) com [mailto:info-ingres-bounces (AT) kettleriverconsulting (DOT) com] On Behalf Of Karl & Betty Schendel
Sent: February 18, 2009 8:25 AM
To: Ingres and related product discussion forum
Subject: Re: [Info-Ingres] Disk problem ?


On Feb 18, 2009, at 7:55 AM, Jürgen Cappel wrote:

The file is shorter than expected. It's probably not a failing disk
drive, but it might have been caused by a system crash in the middle
of filesystem writes. I agree, open an issue if you have support.

Karl


_______________________________________________
Info-Ingres mailing list
Info-Ingres (AT) kettleriverconsulting (DOT) com
http://www.kettleriverconsulting.com...fo/info-ingres

i've also seen similar issues w/ HP-UX 11.0 when running threads. I
believe the
root of the problem was the the thread implementation at the OS level
wasn't
truly POSIX compliant. I was able to continue using threads by
shuffling the
number of consistency points and percentages/intervals in CBF as I
recall.

In my case, the problem would show up at the end of reload.ing, when
adding rules,
procedures and such.

I can't recall if that was on OpenIngres 1.2 or Ingres II 2.0 (too
many years ago
I guess) but I believe it was the later....

The recommended solution at the time was to return it using ii_slaves
and let
handle it, but there was a terrible performance hit