Re: [Info-Ingres] unloaddb and E_US1068 Integer overflowdetected in query

Hi Martin,

This issue has been around for some time (years before my time), now
listed as bug 119782. Since I've been aware of it, I have seen 4 issues
with the same symptoms, all on 2.6/0305 or 2.6/0608 [Whilst I was
writing this Damien Owen also replied - his is one of the issues]. I've
so far tracked it down, via a debug copydb executable, to a duff row
coming back from the esql select on view iitables that is performed by
copydb/unloaddb to get the list of tables to work with. It seems to
occur when the number of tables being handled is large (in the
hundreds), and, at least in some cases, where the tables are being
created and destroyed frequently (eg a bulk load of data from one
installation to another, done at regular intervals via a script).

For one of the current issues I have on this, I was about to build a
patch to include a fix for bug 118096 which fixes some poor concurrency
handling of views that *may* be the root cause of this.

If none of the above sounds applicable to you, I would certainly be
interested in a reproducible testcase on Linux if you can get one
together - we've not been able to reproduce it in-house. I can provide a
debug copydb/unloaddb also.

Regards,

Kevin Smedley,
Ingres.



> -------- Forwarded Message --------
> From: Martin Bowes
> Reply-To: Ingres and related product discussion forum
>
> To: Ingres and related product discussion forum
>
> Cc: eoinogorman-at-gmail.com
> Subject: [Info-Ingres] unloaddb and E_US1068 Integer overflow detected
> in query
> Date: Tue, 26 Aug 2008 13:17:39 +0100
>
> Hi All,
>
>
>
> I've got a situation where unloaddb will occassionally produce an error:
>
> INGRES UNLOADDB Copyright 2003 Ingres Corporation
>
> Unload directory is '/user/ingres/work.II/sharp_dev'.
>
> Reload directory is '/user/ingres/work.II/sharp_dev'.
>
> E_US1068 Integer overflow detected in query.
>
>
>
> The error is not reproducible.
>
>
>
> I have tried both binary and ascii unloads. I have tried recovering the
> database to moments just before the error is produced. And on all
> occassions the unload works perfectly well.
>
>
>
> There is nothing in the errlog to indicate what the problem might be….or
> that a problem occurred.
>
>
>
> I'm running II 2.6/0604 (int.lnx/00) + patch11814.
>
> OS: Red Hat Linux release 7.1 (Seawolf)
>
>
>
> I'm recoding some of the scripts where this problem has been observed to
> run 'set trace point sc924 4200' to see if I get lucky and actually
> capture a query at the next occasion.
>
>
>
> Has anyone had any experience of this problem and could provide a
> pointer or two? I note that cdi has a mention of a similar problem by
> Eoin in 2005, but I couldn't find a resolution to that issue.
>
>
>
> Martin Bowes
>
>
> _______________________________________________
> Info-Ingres mailing list
> Info-Ingres-at-kettleriverconsulting.com
> http://www.kettleriverconsulting.com...fo/info-ingres