Hi,
I'm trying to set up an HA system with PostgreSQL. I have three
computers attached to a SAN system (they share a filesystem).
I have installed PostgreSQL (7.1.2) in all of them, created a DATA
directory in the shared filesystem, 'inited' the database with one of
the systems, and tryed to alternatively start the database with every
machine. It all works fine.
The script that starts pgsql on every machine checks if the pidfile
exists so only one of them really does the job. The machine that then
runs the server alters it's shutdown procedure to stop the database as
clean as possible.
We have been playing with the database to see what happens if two of
them try to run the database at the same time... better don't try it ;-)
All of these things have made me wonder about how nice it would be to
have some sort of clustering in PostgreSQL. Is there any plan for this?
It would be nice to have N instances of PostgreSQL running on N
different machines, allowing clients to connect to any of them, seeing
the same database whatever the server they connect to. Servers would
share the same physical database. It could be by sharing files (via SAN
or NFS) or without file sharing, having some sort of protocol between
servers.
I can imagine the features, but I also imagine that this would take a
lot of effort. Is there any plan for this? anyone else has thought about
it? I'm sure there is.
See ya,
Christian Pomar