--- In firebird-support@yahoogroups.com, Helen Borrie <helebor@t...>
wrote:

> At 09:53 PM 26/04/2004 -0400, you wrote:
> >I'm using FB1.5 (release version) in Classic mode on a new

multiprocessor

> >Windows2003 server with 1GB of ram.
> >
> >The database is 10GB, page size 8192, buffers 2048.
> >
> >I'm running it in test mode right now to see how the Classic version
> >behaves. I just came across this problem in the logs:
> >
> >Internal gds software consistency check (page in use during flush 210))
> >
> >In the 15 seconds prior to this message in the log, there are two 10061
> >connect errors, and then a 'Guardian starting' message.
> >According to the few messages I can find about this error (none of them
> >recent), this kind of message usually is reported along with a corrupt
> >database complaint.
> >
> >Can someone give me a clue as to what this message means, and what

can be

> >done to prevent it?
>
> It gives me a bad feeling. May I suggest you report it into
> firebird-devel?

It will be _very_ helpful if it is possible to make reproducable
test case - database and application. Suspected scenario is: Some
transaction is started and do nothing while ~150 000 other
transactions are subsequently started and made some changes, some
commited/rolled back shortly, some live for a long time. At last first
forgotten transaction tries to read data changed by transaction
started within ~50 000 range after it. One of my friends tries to make
such reproducable test on Linux CS just now.

Best regards,
Alexander