Very Computer

by **Tom Gillesp** » Wed, 11 Dec 1996 04:00:00

I have an ingres database that keeps dying with the errors:

E_DM934D_DM0P_BAD_PAGE Page is not consistent with buffer descriptor:
E_DM9306_DM0P_UNFIX_PAGE Bad page address.

After these errors appear in the errlog.log file, the iidbms server dies.
My question is, can these page errors occur from anything besides a
disk problem ? This has happened many times in the last several weeks,
against different tables, and the tables are on the same disk.

--
_/ _/ Tom Gillespie
_/ _/ Harris Semiconductor
_/ _/ Palm Bay, FL USA

_/ _/ www.harris.com
_/ _/
_/ _/ And the deamon said: rm -r * and then there was nothing.

by **Karl Schend** » Thu, 12 Dec 1996 04:00:00

Quote:> Subject: Error getting pages !
> I have an ingres database that keeps dying with the errors:

> E_DM934D_DM0P_BAD_PAGE Page is not consistent with buffer descriptor:
> E_DM9306_DM0P_UNFIX_PAGE Bad page address.

> After these errors appear in the errlog.log file, the iidbms server dies.
> My question is, can these page errors occur from anything besides a
> disk problem ? ...

What platform and what Ingres version?
I've seen this happen on an Alpha running 6.4 and Digital Unix, and it's
definitely NOT related to a disk error. I have some notion that I've also
seen this one listed in a bug-fix list but I can't remember where (might
have been related to OpenIngres rather than 6.4).

Are you getting multiple bad page addrs, or is this the same one happening
over and over? once a table gets hit like this you have to copy out,
drop, and recreate it. verifydb -otable will fix the table so that you
can copy it out. Do this with Tech Support's help, though, in case they
want you to try something else first.

--
Karl Schendel Phone: (412) 963-8844
Telesis Computer Corp Fax: (412) 963-1373

by **Malcolm Baldw** » Tue, 17 Dec 1996 04:00:00

> I have an ingres database that keeps dying with the errors:

> E_DM934D_DM0P_BAD_PAGE Page is not consistent with buffer descriptor:
> E_DM9306_DM0P_UNFIX_PAGE Bad page address.

> After these errors appear in the errlog.log file, the iidbms server dies.
> My question is, can these page errors occur from anything besides a
> disk problem ? This has happened many times in the last several weeks,
> against different tables, and the tables are on the same disk.

We have been suffering these problems repeatedly on 6.4/06 on Solaris.
Our system is down at this very minute due to one of these crashes which
has corrupted the table (sometimes just starting the server seems to fix
it). I believe it is related to a Btree corruption problem but so far
Tech support have been unable to reproduce the problem. We have checked
our disks and they are fine. Please play hell with Tech support about
this and we may get it resolved.

--

MIS Tel +44 (0)1524 593320
Lancaster University Fax +44 (0)1524 844011
LA1 4YW ENGLAND

by **PRIYES** » Mon, 23 Dec 1996 04:00:00

I have encountered these types of errors on OpenIngres 1.2 (NT and Sun
Solaris).

The problem is caused by corruption of data pages in the DMF cache and
there
are probably a number of causes. There are some precautions you can take :

a). Don't use compression for tables that are updated - note that tables
generated
by create table as select .. will default to compressed heap.

b). When this happens, shutdown ingres and check that shared memory is
cleared. Shared memory not clearing when ingres is shutdown and restarted
may be one of the causes of this happening.

c). When a problem like this occurrs, get users off and check the table
and associated indexes. eg. select * from .. note some tests like verifydb
may indicate that the table is OK but there may be a corrupt page lurking.

Priyesh.

Mercurie Software Consultants