Hello -
I am curious as to people's opinion about best practices in deployment of
pgsql in a high-availablity (HA) situation. By HA I mean that data loss
must be as close to zero as possible, and that web servers using the DB
need to be deployed in multiple geographic locations world-wide, to
provide for network availability.
In my mind, this suggests a need for a multi-master configuration, such as
that which can be achieved with pgcluster. I see redundency both within a
POP, as well as cross-POP. I am however concerned about the latency
implied by POP locations (let's assume a POP in the EU, US and JP).
I am aware of bizgress, though am concerned that it doesn't really address
the wide-area replication issue. It does address the HA aspect within a
single POP quite nicely though.
How do you deal with this, commerical or OSS?
Thanks in advance for suggestions.
- Marc
