I was thinking about that (as per your presentation last week) but my problem is that when I'm building up a series of inserts, if one of them fails (very likely in this case due to a unique_violation) I have to rollback the entire commit. I asked about this in the novice<http://postgresql.1045698.n5.nabble.com/execute-many-for-each-commit-td5494218.html>forum and was advised to use SAVEPOINTs. That seems a little clunky to me but may be the best way. Would it be realistic to expect this to increase performance by ten-fold?
On Mon, Feb 20, 2012 at 3:30 PM, Josh Berkus <[email protected]> wrote: > On 2/20/12 2:06 PM, Alessandro Gagliardi wrote: > > . But first I just want to know if people > > think that this might be a viable solution or if I'm barking up the wrong > > tree. > > Batching is usually helpful for inserts, especially if there's a unique > key on a very large table involved. > > I suggest also making the buffer table UNLOGGED, if you can afford to. > > -- > Josh Berkus > PostgreSQL Experts Inc. > http://pgexperts.com > > -- > Sent via pgsql-performance mailing list ([email protected]) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-performance >
