On Tue, 2005-03-29 at 17:34, David B wrote:
> Folks,
>
> I'm thinking of replication tools for high volume site.
> Postgres on dual Opteron box. 1/2 GB of data.
>
> Big overnight load of data.
> Some inserts during the data but mostly selects.
>
> I think I've narrowed it down to Slony and Mammoth.
> Any pro's or con's on one vs the other?
>
> There is the obvious price difference but thats managable if Mammoth
> substancially better.
>
> Some area where our environement might stretch a solution.
> Some records can be large. Limitied to one table but they can be 40K+
> of chraracters in a column.
>
> High volume of inserts coming during nightly load (millions of rows).
> These loads are coming in commited blocks of 100.
>
> That cannot slow down the replication.
> Our call center agents needs access to them asap.
>
> Any locking issues on the master during the replication process?
>
>
> Anything else I should be concerned about or things I need to consider?
> Personal experiences you have had running either.
I don't see any obvious problems for either replication engine here.
Since slony is free and pretty easy to setup, I'd make a test setup and
see how it handles the load.
Generally speaking I'd expect Mammoth to handle high loads a little
better since it's a log shipper and not trigger based. So, if slony
proves too slow for this situation, then I'd recommend asking the folks
at Command Prompt for a demo of their system to see if it can.