Hi all,

I have a postgres DB where beside a load of common tables, each new
customer gets one special table of it's own. Customers are created quite
frequently by our admins.

The DB is replicated to a slave via slony. The problem is that once the
slony replica is set up, the new tables have to be added to the
replication each time we create a new customer on the system. This poses
a problem, as the DB is quite heavily loaded usually, and the locks
involved are problematic.

Now all of these per customer tables are not actually very often
accessed, so if I could take them out to a separate replication set I
guess it would not be anymore so problematic to include adding the per
customer table to replication in the automated customer creation script.

All these per customer tables have a foreign key to a central 'common
denominator' table, which is the source of the biggest contention on the
system. Locking this table for 2 minutes brings our system to a halt
(due to all connections of our connection pool ending up waiting on the
lock)... so I was thinking putting the central table (together with all
the other common tables) into a separate replication set, and the per
customer tables on a second replication set.

The question is, will it work properly, having foreign keys from one
set's tables to the other set's tables ?

TIA,
Csaba.


_______________________________________________
Slony1-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/slony1-general

Reply via email to