December 12th, 2013, 11:30 AM
Which replication - going from 8.4/slony to 9.1
I'm building 2 new debian Wheezy servers which run postgres 9.1.
Previously we used 8.4 and Slony.
One server is the main server (web server as well), the other is a backup on another site.
Both servers have their own separate data feeds, but some data (e.g. user data) is replicated from main to backup. In the event of a problem with the main server, users can use the backup (but with no write access to the replicated tables). A pretty normal setup I'd guess.
So, given that I find slony a bit of a black art , should I be using the the built in replication now available, or wil slony suit my setup better?
January 9th, 2014, 11:55 PM
I'm not sure I understand "Both servers have their own separate data feeds". What does that mean? Is there anything doing inserts and updates directly to the secondary server, or is everything from server A replicated to B?
Built-in streaming replication in 9.1 is very easy to set up and manage, but it's an all-or-nothing deal. You can't just choose to replicate some parts of the database, as with Slony.
If you want fast, complete single-master replication with minimal difficulty and minimal downtime, choose built-in streaming. If you need partial replication or if you have additional logic to be handled in replication, then you will have to use Slony, Londiste, or one of the other trigger-based replication systems.
January 10th, 2014, 05:28 AM
> Is there anything doing inserts and updates directly to the secondary server,
Yes, on some tables. Looks like it's slony then, cheers.