On Wed, 1 May 2002, Tom Lane wrote:
> Have you tried running frequent (more than hourly) non-full vacuums?
I'd love to. But one of our customers is keeping us from upgrading to
7.2. (the bastards.) Full vacuums are my only choice for the time
being. But knowing how vacuum works, how would this help? It would
keep our queries nice and speedy, but unless Postgres has a new, magical
way of knowing outdated versions of a row are actually outdated, the
overall file bloating will be the same, and we'd still need the full
vacuums every hour.
If vacuum in 7.2 marks the row as outdated, and new data can be
placed *over* it, or a full vacuum completely obliterates *all* rows
marked as invalid, sure. But if it keeps some kind of list of all
recent row versions to avoid a sequence scan through all the invalid
rows, should that list grow large, I could see it losing track with
our large turnover rate.
What to do...
--
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+
| Shaun M. Thomas INN Database Administrator |
| Phone: (309) 743-0812 Fax : (309) 743-0830 |
| Email: sthomas@townnews.com AIM : trifthen |
| Web : www.townnews.com |
| |
| "Most of our lives are about proving something, either to |
| ourselves or to someone else." |
| -- Anonymous |
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+