On 10.8.2012 21:46, Helen Borrie wrote:

>
> At 04:48 AM 11/08/2012, Lester Caine wrote:
>
> >I had thought that caching of write data in windows had been sorted,
> but I'm not
> >entirely surprised if just some block of data has been lost.
>
> I believe the OP mentioned that they were using Fb 2.1.3. That's
> almost three years old now and two sub-releases out of date. I suspect
> the moves to improve the caching behaviour on Windows came more recently.
>
> >If it was one of my sites I'd drop a new machine in just to prove a
> point ...
>
> The system log reports of disk corruption are suspicious...but they
> don't account for lost generator values. If the generator page was
> actually written to a corrupt block, I'd expect the page to be kaput
> and to see other evidence of corruption. It seems more likely the page
> write was lost when they had that power failure.
>

This makes sense to me, can I check if generator page was lost if it was
written to corrupt block? There is no indication of database corruption
within firebird.log

>
>
> I wonder whether anyone has checked to see whether Forced Writes is
> off. If so, the chances of pending writes disappearing on a power
> outage would be high.
>
> ./heLen
>

Forced writes are on. That was among first things I checked.

>
>

[Non-text portions of this message have been removed]