Archive for May 30th, 2006
One of the greatest WTF moments I have had in recent times in relation to technology was the time when I discovered that Postgresql did not have decent replication built into it. Transactions and ACID compliance are good things to have, but any self-respecting RDBMS solution that calls itself enterprise-ready should also have native replication capabilities.
From what I have seen, for us (using the 8.x branch of Postgresql), the available options are the commercially available Command Prompt’s replication solution, Slony I, PGCluster and a whole host of middleware applications available only if you are using JDBC.
Of the lot, Command Prompt has not bothered to get back to me with a quote for their product (very few companies actually reply to such queries, weird) and from what I had seen mentioned somewhere it costs somewhere in the region of USD1000.
Slony I looks pretty okay and it has been around for a while too, but it does not support schema replication, leaving you high and dry unless you are lucky enough to work somewhere the database schema does not change for the duration of your employment. Moreover, it does not have multimaster capabilities.
PGcluster seems to have most of what we want in a single solution — clustering, load balancing and failover.