| From: | Simon Riggs <simon(at)2ndquadrant(dot)com> |
|---|---|
| To: | Dan Harris <fbsd(at)drivefaster(dot)net> |
| Cc: | pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: Quad Opteron stuck in the mud |
| Date: | 2005-07-13 22:51:16 |
| Message-ID: | 1121295076.3970.381.camel@localhost.localdomain |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
On Wed, 2005-07-13 at 12:54 -0600, Dan Harris wrote:
> For example, as I'm writing this, I am running an UPDATE statement
> that will affect a small part of the table, and is querying on an
> indexed boolean field.
An indexed boolean field?
Hopefully, ftindex is false for very few rows of the table?
Try changing the ftindex to be a partial index, so only index the false
values. Or don't index it at all.
Split the table up into smaller pieces.
Don't use an UPDATE statement. Keep a second table, and insert records
into it when you would have updated previously. If a row is not found,
you know that it has ftindex=true. That way, you'll never have row
versions building up in the main table, which you'll still get even if
you VACUUM.
Best Regards, Simon Riggs
| From | Date | Subject | |
|---|---|---|---|
| Next Message | jobapply | 2005-07-13 23:46:23 | Indexing Function called on VACUUM and sorting ? |
| Previous Message | jobapply | 2005-07-13 22:35:10 | Functional index is 5 times slower than the basic one |