> The corruptions happen with at FB 1.5.3 SuperServer on Windows 2003
> Enterprise (with SP2 and all current MS Updates). The DB file is

stored on

> SAN (HP HSV210 Multi-Path) attached with Fibre Channel HBAs. The DB

has

> forced writes enabled.
>
> The firebird.log shows:
>
> internal gds software consistency check (error during savepoint

backout

> (290))
>
> before the DB craps. This message is also displayed on one of the

attached

> client connections at the time it is generated.
>
> A subsequent gbak run on this DB terminates with:
>
> "database file appears corrupt ()
> -bad checksum
> -checksum error on database page 4149836"

This is not good -- although the pages have a checksum value, the
checksum logic has been inactive for a LONG time in FB or IB. The
current checksum is a constant "12345" value.

This is a sign on a major hardware failure, pages are not being written
correctly to disk.

> This DB has been running fine for a while now before showing this

strange

> behavior. Even stranger was that after starting with new DB file

created

> on that system we run into the same error within a week, down to days,
> from using the new DB file.
>
> On top of that we do not use SQL save points in any of your DML which

lead

> to the questions:
>
> - why do we get this message not using save points ?

The engine has its own internal savepoints, which are distinct from the
SQL "savepoints", so this can be confusing.

> - what might be a possible cause of this message ?
> - where do we most likely have to look to fix/prevent this ?

The only thing that I can recommend is to move the database to different
storage device, to see if the cause is the SAN.


Sean