[comp.databases] Major data corruption in INGRES

manatt@lll-winken.LLNL.GOV (Doug Manatt) (04/01/89)

     For the last 1.5 years we have been having problems with the
reliability of the INGRES (rti, commercial) rdbms system.  The
symptoms we have been experiencing are major corruption of the
"backend" "system" tables and also corruption of index pages in
btree table structures.  

     I have tried to work this out with the rti tech support
team, but so far, the response that I receive is that they are
unable to duplicate my problems at their site.  

     What I need to know from the net is whether others have
similar experiences with their rdbms packages, and in particular
what sort of support has been provided in similar situations.  We
are reevaluating our use of ingres and if we can find a more
reliable package we might think of switching.

                         Doug Manatt
                         manatt@lll-winken.llnl.gov
                         (415) 422-7257

staff_bob@gsbacd.uchicago.edu (04/01/89)

> 
>     For the last 1.5 years we have been having problems with the
>reliability of the INGRES (rti, commercial) rdbms system.  The
>symptoms we have been experiencing are major corruption of the
>"backend" "system" tables and also corruption of index pages in
>btree table structures.  
> 
For what little this is worth, we've had very little trouble with Ingres
data integrity. I've seen a couple of cases of corrupt BTREES (symptom:
inconsistant query results), and the best RTI could do was show us how
to fix them. No one was ever able to identify when and how they occurred.
We've never had any corruption of any of our system tables, to my knowledge.
All this in a VMS cluster, primarily running VMS Version 4.7.

I seem to recall some comment to the effect that 'we do see these BTREE
problems occassionally, but we haven't been able to figure out what's
causing them'. At any rate, that is obviuosly true. I would venture to say
that if your site has this problem repeatedly, RTI should be very interested
in using your site to isolate the problem. Try going up the Tech Support
ladder a couple of notches...

R.Kohout

Jeff_A_Scott@cup.portal.com (04/03/89)

> 
>     For the last 1.5 years we have been having problems with the
>reliability of the INGRES (rti, commercial) rdbms system.  The
>symptoms we have been experiencing are major corruption of the
>"backend" "system" tables and also corruption of index pages in
>btree table structures.  

We've been using INGRES for about 5 years now and have been through
all of the usual and unusual problems with them.  Don't know what version
you're on but you might try turning the local page cache off.  RTI has
had some problems with the local/global caching among multiple
database accessors.
				Jeff Scott
				jscott@cup.portal.com
				... !sun!portal!cup.portal.com!jscott

bg0l+@andrew.cmu.edu (Bruce E. Golightly) (04/03/89)

That's interesting. We've been using Ingres for 3+ yeats, and have not
had many problems in this area. Usually when we have, we can trace the
source of the problem to someone doing something silly, like shooting
a backend.

I would like to hear more about this, and expect others in the Ingres
community would, too. What are the circumstances? What environment?

Bruce

brunjes@isctsse.UUCP (Roy Brunjes) (04/05/89)

In article <22812@lll-winken.LLNL.GOV> manatt@lll-winken.LLNL.GOV (Doug Manatt) writes:
>
>     For the last 1.5 years we have been having problems with the
>reliability of the INGRES (rti, commercial) rdbms system.  The
>symptoms we have been experiencing are major corruption of the
>"backend" "system" tables and also corruption of index pages in
>btree table structures.  
>
>     I have tried to work this out with the rti tech support
>team, but so far, the response that I receive is that they are
>unable to duplicate my problems at their site.  
>
>     What I need to know from the net is whether others have
>similar experiences with their rdbms packages, and in particular
>what sort of support has been provided in similar situations.  We
>are reevaluating our use of ingres and if we can find a more
>reliable package we might think of switching.
>
>                         Doug Manatt
>                         manatt@lll-winken.llnl.gov
>                         (415) 422-7257

Sorry for posting this ... the email bounced (again)  8^(

Doug -- I have used Ingres since 1985 and have found it to be very reliable
for the project I have been working on (and in fact have heard of no
such problems in Kodak).  We have hit bumps in the road with Ingres
(no vendor's products excludes you from that pleasure!), but the RTI
folks have been very helpful.  We hit a problem that RTI couldn't
reproduce either, and in that case, they dialed in and found the
problem in a little over one day.  It turned out to be their problem
and they fixed it right away -- they had to cut a tape and Federal Express
it, but we were running in the meantime (limping more like it) and
within 2 days, things were fine.  Sorry that I can't say I have heard
of a similar problem here and we did xyz to fix it.  Your nearest RTI
sales person might be the one to turn up the heat in RTI to find the
problem.  Ours did that for us on one item (Sun OS 4.0 support) for
a Sun 4.  Does he/she understand the severity of the situation?

Roy Brunjes   If you dare try email:  ...rochester!kodak!kadsma!brunjes

Standard disclaimer applies:  My opinions, my thoughts, company in dark
about my thoughts and opinions, etc.