[Bucardo-general] What to do when one machine in a multi-master setup fails and a replacement is needed ?

[Dave Garry]

Hi,

We use Bucardo in a multi-master setup on two different machines,
and it seems to be working nicely.

However we need to be able to recover quickly if a machine fails,
by installing a new one, and telling Bucardo to use the replacement
instead of the original, failed, machine.

We call our pair of database machines "box A" and "box B".

If box B fails and is replaced we want to import our schema, then
tell box A to sync to the new box B.

If box A fails, I imagine it would be more tricky as Bucardo
is set up on that box. In this case we would also need to
import our schema, then do an initial sync from box B to box A,
before telling Bucardo to resume normal operation with the new
box A.

Our initial set up is fully scripted / automated, and I'm keen
for the above to be done the same way, such that failed machines
can be automatically rebuilt without the end user having to use
the command line. (I already auto install the OS, our software etc.
in an automated way leveraging PXE and AutoYast (openSUSE)).

Is this scenario documented anywhere ? Does anyone have an example
recipe I can follow to set this up ?

Thanks.

--

Dave G