On Tuesday 30 May 2006 21:13, Brendan Duddridge wrote:
I apologize if this is the wrong list to post these questions.
There is a Slony list, but I'm sure there's some overlap
1. Does Slony replicate table statistics to the slaves? No.
2. Do you still have to vacuum analyze the slaves?
Yes. All a subscriber basically is, is an otherwise normal PostgreSQL
database that a Slony process happens to write data into.
3. Are selects replicated or just inserts, updates, and deletes?
Replication actions are scheduled on a row-level basis, via trigger, for
each row inserted, updated, or deleted. Those actions are executed in an
intelligent manner on the subscribers. Additional events, such as DDL
modification scripts, can also be scheduled.
However, to answer the question, just inserts, updates, and deletes.
4. I'd like to have our updates go to the master and selects go to
the clients. Is that possible?
Yep ... while keeping in mind that it is an asynchronous replication system,
and the subscribers will not generally be in perfect sync with the master.
You will need separate database connections since the setup doesn't appear
as a single cluster or anything like that.
5. What happens if an update tries to go to the slave? Does it
automatically get routed to the master?
Tables on the subscribers are modified to prevent write ops. The subscriber
table otherwise has no knowledge of the master. Operations are therefore
I would strongly suggest setting up a test replication setup to learn
exactly how Slony works before deciding on a particular architecture.
"When a politician says he wants you to serve a cause greater than your
own self-interest, the cause he has is mind is his own reelection."