also something interesting

if i do (wee yyy a completely new value)
delete ... where xxx= yyy
Insert ... (xxx= yyy)
commit => 650 ms (isql show read 2, write 297)

and i do again (exactly the same as previous)
delete ... where xxx= yyy
Insert ... (xxx= yyy)
commit => 600 ms (isql show read 2, write 294)

and i do again (exactly the same as previous)
delete ... where xxx= yyy
Insert ... (xxx= yyy)
commit => 40 ms (isql show read 2, write 148)

and it's always like this, at the 3th time (only) it's go fastly !
why for the exact same intructions the commit need to write more in The 1rt and 2nd call than in the Third call ?

thanks be advance to help me in the firebird mistery

stephane

--- In firebird-support@yahoogroups.com, "hvlad" <hvlad@...> wrote:
>
> > about the index corruption, i deployed on our production server the FB 2.1.4 only 2 days ago (after another FB_inet_server massive hanging up). for now it's seam to work OK (around 200 fb_inet_server.exe running simultaneous), and if i see any probleme appear again i will come back to you !
>
> Just to be sure - you used my custom build, correct ?
>
> If yes, please, look at firebird.log - is it have messages such as
>
> "BTR/garbage_collect : left page is released." ?
>
> Regards,
> Vlad
>