Subject Re: [firebird-support] database corruption without system restart
Author Michal Rzewuski

thank you for answer and sorry i'm going to ask new questions
:( however there may be other people interested in reading
about the stuff that was not discussed on the group (at least
i couldn't find).

so the history of changes looks as follow. the last stat
is taken after sweep process is finished, the first is
made on fresh copy of database file, just after restore.

data pages/data page slots

after restore i deleted some milion of records and after
that i started sweep process (before crash i had about
100e6 and i deleted about 20e6).

the other thing i'm warring about is the OIT, it's not
changing and difference is huge - normaly in the past
the transaction number was a small step behind OAT.

Oldest transaction 25367
Oldest active 244699
Oldest snapshot 244699
Next transaction 257758

the problem is that it has to be working because of consumers
that start to be very nervous. i'm waiting for the weekend
to backup/restore but it has to be working for next 24h :(
every day i get about 0.5 - 1e6 records.

and my questions:
-do new pages for fresh data are allocated from data page slots?
(i mean are data page slots reused? why this space was not
given to the database for any purpose and is always attached
to the same table?).
-should i start sweep process again to reduce the OIT-OAT
difference or do anything else?
-how much space do i have in reserve before next database
-is it a good idea to turn on the 100% filling od data pages?

the average size of record i can calculate:
4424805*8192/100e6=360 baytes (but in real *0.8 because of
80% of filling of data pages).