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 almost three years old now and two sub-releases out of date.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]
