Re: autovacuum next steps, take 2

Jim C. Nasby wrote:

> That's why I'm thinking it would be best to keep the maximum size of
> stuff for the second worker small. It probably also makes sense to tie
> it to time and not size, since the key factor is that you want it to hit
> the high-update tables every X number of seconds.
>
> If we wanted to get fancy, we could factor in how far over the vacuum
> threshold a table is, so even if the table is on the larger size, if
> it's way over the threshold the second vacuum will hit it.

Ok, I think we may be actually getting somewhere.

I propose to have two different algorithms for choosing the tables to
work on. The worker would behave differently, depending on whether
there is one or more workers on the database already or not.

The first algorithm is the plain threshold equation stuff we use today.
If a worker connects and determines that no other worker is in the
database, it uses the "plain worker" mode. A worker in this mode would
examine pgstats, determine what tables to vacuum/analyze, sort them by
size (smaller to larger), and goes about its work. This kind of worker
can take a long time to vacuum the whole database -- we don't impose any
time limit or table size limit to what it can do.

The second mode is the "hot table worker" mode, enabled when the worker
detects that there's already a worker in the database. In this mode,
the worker is limited to those tables that can be vacuumed in less than
autovacuum_naptime, so large tables are not considered. Because of
this, it'll generally not compete with the first mode above -- the
tables in plain worker were sorted by size, so the small tables were
among the first vacuumed by the plain worker. The estimated time to
vacuum may be calculated according to autovacuum_vacuum_delay settings,
assuming that all pages constitute cache misses.

--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support