On Wed, Nov 11, 2009 at 09:35:35AM -0800, Ben Chobot wrote:
- What are you trying to protect against? Software failure? Hardware
- failure? Both?
-
- Depending on your budget, you could theoretically point any number of
- failover nodes at a san, so long as you make sure only one of them is
- running postgres at a time. Of course, you still have the single point
- of failure in the SAN. If you aren't made of money and are running
- linux, we've found DRBD is a great way to cluster two machines and it
- avoids a few single points of failure. But you limit yourself to two or
- three cluster nodes.
Protecting against both hardware and software failure.
SAN failure would be handled by the offsite node, but we've got a pretty robust
SAN, (I don't have all of the details) so it may even not have a single point
of failure.
We tried out DRBD and the performance impact was pretty sigificant. our
app is very sensitive to any performance hitch so I just can't see any
form of replication working for us.
Dave