The clone machine was previously the main server.
It is formatted ext3.
It suffered corruption 3-4 times.

My short list of potential causes are down to these variables.

* 64 bit Linux distributions.
* Legacy Windows applications using the Firebird ODBC driver.
* An obscure bug in Firebird.

or whatever causes the following message about 5-10 times a day.

"INET/inet_error: read errno = 9"


Again my issue is not that Firebird corrupts, ANY database can do
that, rather that the nil data loss recovery process is so difficult.

Thanks again.

--- In firebird-support@yahoogroups.com, Milan Babuskov <milanb@...>
wrote:

>
> derryck.welas wrote:
> > see your using XFS.
> > Try ext3 , used it for more than 3 years without a glitch and running
> > on fb2.0.3 (preparing for 2.1)
>
> Agreed. Because of the way they internally work, databases on
> filesystems like XFS and ReiserFS are more likely to get corrupt in

case

> of the power failure or system crash (kernel crash). Ext3 is almost
> unbreakable, even with forced writes off (because of the way it writes
> data to the disk). Even better alternative would be to skip the
> filesystem completely and use the raw partition (but that has some

other

> drawbacks). I plan to talk in more detail about all this at this year's
> Firebird Conference.
>
> Of course, if OP didn't have power failure or system crashes, something
> else might be the problem.
>
> --
> Milan Babuskov
> http://www.guacosoft.com
>