[rt-users] RT Scalability

[Mark Chappell]

Kenneth Marshall wrote:
> I would love to see a few more details of your PGCLuster
> setup since we are in the process of upgrading to the 3.6 release on
> replacement hardware and would like to increase our redundancy, if
> possible. 

PG Cluster is basically a shared nothing architecture, although they're 
  spinning off PG Cluster II which would have a shared backing store. 
However we seem to be so read heavy that the write speed benefits of a 
shared backing store aren't really needed.

2 machines dedicated to RT.

Basic layout.
3 replicators, 1 on a totally separate box so that quorum can be 
attained if one of the 'RT' machines disappears, and the other 2 running 
on the 2 'RT' machines.

2 cluster database nodes, one on each of the 'RT' machines.

Apache is running on each of the 'RT' machines, and queries the database 
node on that machine.

We're contemplating rolling out another pair of DB nodes, and then using 
the pglb part of PG Cluster, but for now we're not using them.

> Since you are using PostgreSQL, have you looked at constraint
> exclusion to partition your tables on the backend. Also, if you recluster
> the tables appropriately, the older information will automatically be
> segregated from the new active tickets -- just some ideas.

And some of those magic words that make searching for useful information 
so much simpler start turning up.  Thank you.


Mark
-- 
With sufficient thrust, pigs fly just fine. However, this is not
necessarily a good idea. It is hard to be sure where they are going to
land, and it could be dangerous sitting under them as they fly
overhead.
  -- RFC 1925