On Sat, Feb 14, 2009 at 12:35 AM, alex <[email protected]> wrote: > Oh, I was thinking in the possibility of active pgpools in all nodes.
That would result in replication problems and information mismatch. pgpool-II is not, AFAIK, ready for active-active clusters, but for active-pasive clusters. Of course, wait for Tatsuo's reply as he will be the one to confirm my thoughts. > This way, each app server talks to a pgpool instance on the same node, > with no load balancing- load balancing is taken care of via app server > load balancing. > > Ascii drawing: > > /--> App --> PgPool ---> PG > | | > | | Write queries > | | > LB +--> App --> PgPool --> PG > | | > | | Write queries > | | > \--> App --> PgPool --> PG > > > Read queries incur no network latency, app is load balanced. Nodes are > homogeneous. You can prevent this from happening by setting load balance off but, of course, you will only be using one node. > Thoughts? You are trying to get the best from an active-active cluster and an active-passive cluster without any of the downsides, which is, AFAIK, impossible :) -- Jaume Sabater http://linuxsilo.net/ "Ubi sapientas ibi libertas" _______________________________________________ Pgpool-general mailing list [email protected] http://pgfoundry.org/mailman/listinfo/pgpool-general
