Re: suggestions/hints for multimaster solution

2009/1/9 Gerd König <koenig(at)transporeon(dot)com>:

> we want to have a scenario where (at least) 2 servers are r/w accessible
> and if one of the server crashes this should cause (almost) no downtime
> for the database.
> Yes, you're right, it's more like "keeping two databases in sync".

I would call this replication with failover (active-passive). To me,
multimaster sounds like active-active, where clients can send queries
to any of the nodes (that is why all nodes are masters). Of course,
it's just my point of view. :)

> We also use heartbeat for pinging the server themselves. Is it also
> possible to use heartbeat for checking the availability of a service
> (I'm not that familiar with heartbeat, so long) ?
> Does this mean you have two identical installations of pgpool-II on each
> of the db-servers ?

Yes, I have two almost identical installations of pgpool-II on each of
the PostgreSQL servers. pgpool-II handles the pgsql servers. Heartbeat
provides HA to pgpool-II.

Hope it helps, at least clarifying. :)

--
Jaume Sabater
http://linuxsilo.net/

"Ubi sapientas ibi libertas"