[PostgreSQL-Hackers] Vacuum and Transactions

	Gaetano Mendola
	Oct 5, 2005 at 12:19 pm

Rod Taylor wrote:
I have maintenace_work_mem set to about 1GB in size.

Isn't a bit too much ?

Regards
Gaetano Mendola

Search Discussions

Andreas Zeugswetter: periodically If we started the vacuum with the indexes, remembered a lowest xid per index, we could then vacuum the heap up to the lowest of those xids, no ? We could then also vacuum each index separately. Andreas

Follow ups

Rod Taylor: Understood. I've seen them, but until they're well tested in the newest version I won't be using them in a production environment. I do appreciate the goal and look forward to this concept being applied or a method of splitting up the work vacuum needs to do, in the future. --
Rod Taylor: Accomplishing the pg_listener cleanup often enough can be difficult in some circumstances. I have other items in this database with high churn as well. Slony was just an example. --
Chris Browne: Yeah, I want to do some more testing of that; it should be easy to improve the "abuse" of pg_listener a whole lot. That eliminates the ability to utilize transactions on things that ought to be updated in a single transaction... -- output = ("cbbrowne" "@" "ntlug.org") http://cbbrowne.com/info/lsf.html MS-Windows: Proof that P.T. Barnum was correct.

Discussion Navigation

view	thread \| post
posts	‹ prev \| 9 of 12 \| next ›

Discussion Overview

font	variable	Hotkey: f
user style	avatars	Hotkey: a