Ann W. Harrison írta:

>
> Stranger is that the page number is 24. The first pages of
> a database hold system table information and they're created
> with the database. I think page 24 is normally a pointer page
> for the RDB$FILES system table.

RDB$PAGE_TYPE is 4 and RDB$RELATION_NAME is RDB$FILES

>
> Did the error happen immediately after the backup started? Gbak
> reads the system tables first, so if the error wasn't essentially
> instantaneous, it had no business looking at page 24.

Sorry Ann. I forgot to write, we use Services API for do backup.
Fortunately I am paranoid and always do backup with verbose.

Log of a normal backup:
.
.
.
gbak: writing domain RDB$1440
gbak: writing domain RDB$1441
gbak: writing domain RDB$1442
gbak: writing domain RDB$1443
gbak: writing domain RDB$1444
gbak: writing domain RDB$1445
gbak: writing domain RDB$1446
gbak: writing domain RDB$1447
gbak: writing domain RDB$1448
gbak: writing domain RDB$1449
gbak:writing shadow files
gbak:writing character sets
gbak:writing collations
gbak:writing functions
gbak:writing tables
.
.
.

Log of the problematic backup:
.
.
.
gbak: writing domain RDB$1440
gbak: writing domain RDB$1441
gbak: writing domain RDB$1442
gbak: writing domain RDB$1443
gbak: writing domain RDB$1444
gbak: writing domain RDB$1445
gbak: writing domain RDB$1446
gbak: writing domain RDB$1447
gbak: writing domain RDB$1448

(End of log file)

>
>
>
> Best regards,
>
> Ann

Gabor