Re: how to failover database connection

~V~ wrote:

> I have two same database client running, when first client crashes, the
> other client will replace it automatically, but I don't know how the 2nd
> client picks up the database connections created by the 1st client.

Not really possible. For this to work you must first have a failover network connection.

Can you continue a FTP TCP session by another process, when the session owning that session, crashes?

Think carefully of something like seq numbers in a TCP connection and a 3rd party taking over the connection.

Not that it is impossible, as ther are some very sophisticated attacks on trusted computers which employ similar techniques.. But doing this as a robust implementation in a production environment...?

Not bloody likely, unless the server software itself (Oracle in this case) provides you with the means of a new session inheriting the session id, shadow process (if running in dedicated server mode) and PGA/UGA of a session that may no longer exist.

If you have long running processes (batch jobs), these should be run *within* the database. Use PL/SQL or Java or even external C stored procs to code the jobs. Schedule them with DBMS_JOB.

Running such processes from a client program is not the best of ideas - unless you can afford a job terminating after 30 hours when it is 99% complete, and then having to undo and redo the complete process.

A client should simply start the process in the database.

--
Billy