[Bucardo-general] First use of Bucardo in production

[Ed Phillips]

Hello all,

I'm new to bucardo. I've chosen it as a flexible and reliable way to
replicate many but not all tables in our production postgres dbs.

The first use case will be the master/slave replication of many but
not all of the tables in one of our production dbs.

It looks like it is not a bad practice, especially with a dedicated
slave server with nothing else on it, to have the bucardo db and the
MCP on the server that also hosts the slave? And not on the server that
has the master?

There could be use cases where we might want to run one central MCP that replicates
a number of remote masters to separate local slaves.

I'm interested at this early stage in learning what the best practices
are for the use and setup of Bucardo. From perusing the list, I see
that running the latest version from the github master is probably
best. At least until version 5 is formally released.

So, to sum up with direct questions:

What are the best practices for setup and install of bucardo for
master/slave?

Are there issues with putting the MCP and the bucardo db on one of the
slaves in single master, multiple slave scenario?

What are the best practices for preparing a slave to become a master
in the event of a catastrophic failure of the master db?

And finally, are there some helpful hints on failover planning? Justin
Tolley, for example, mentioned a pgdump of the bucardo db on every update to the
syncs, as a part of a minimal contingency plan for ensuring that
bucardo survives a failure of the node that has the bucardo db
and the MCP.

Moving to bucardo, is already moving in the right direction. I want to
to do the move to bucardo in as wise and informed a way as
possible. Hence, these general questions.

Thank you.