I am planning a system that will be hosted within a secure network with very
limited access. All access to the database will be from within this network.
I need to have a copy of the database available on a much less secure
server, which will allow access for web applications. This copy of the
database is considered sacrificial.

The replication methods I have read about are focussed on creating standby
servers to take over if the primary becomes unavailable. For this, there is
obviously a requirement that the slave be able to connect to the database on
the master. This is not allowed by our security set-up (that's the whole
reason for having the replica database).

I need to somehow push the changes from the master out to the slave. The
database is not very complex or large (two main tables with < 100,000 rows)
and new rows will be added only every few minutes. A minute or so lag on the
copy database is acceptable.

I think Slony-1 using shipping of log files may be an option:
http://www.slony.info/documentation/1.2/logshipping.html

Could I do this without using a separate replication tool? Is there any way
that I could use the replication methods in PostgreSQL 9 without allowing
the slave to connect to the master?

I know that log shipping is still possible with earlier versions of
Postgres, but I think the slave is then in a permanent state of recovery and
cannot accept read-only connections.

Thanks in advance for any suggestions,
Alys

Reply via email to