Hi Kevin, comments after your comments
2011/12/3 Kevin Grittner kevin.gritt...@wicourts.gov:
Ernesto Quiñones wrote:
Scott Marlowe wrote:
Ernesto Quiñones wrote:
I want to know if it's possible to predict (calculate), how long
a VACUUM FULL process will consume in a table?
I don't think you said what version of PostgreSQL you're using.
VACUUM FULL prior to version 9.0 is not recommended for most
situations, and can take days or weeks to complete where other
methods of achieving the same end may take hours. If you have
autovacuum properly configured, you will probably never need to run
VACUUM FULL.
I'm working with PostgreSQL 8.3 running in Solaris 10, my autovacuum
paramaters are:
autovacuum on
autovacuum_analyze_scale_factor 0,5
autovacuum_analyze_threshold5
autovacuum_freeze_max_age 2
autovacuum_max_workers 3
autovacuum_naptime 1h
autovacuum_vacuum_cost_delay -1
autovacuum_vacuum_cost_limit-1
autovacuum_vacuum_scale_factor 0,5
autovacuum_vacuum_threshold 5
my vacuums parameters are:
vacuum_cost_delay 1s
vacuum_cost_limit 200
vacuum_cost_page_dirty 20
vacuum_cost_page_hit1
vacuum_cost_page_miss 10
vacuum_freeze_min_age 1
Ah, well that right there is likely to put you into a position where
you need to do painful extraordinary cleanup like VACUUM FULL. In
most situation the autovacuum defaults are pretty good. Where they
need to be adjusted, the normal things which are actually beneficial
are to change the thresholds to allow more aggressive cleanup or (on
low-powered hardware) to adjust the cost ratios so that performance
is less affected by the autovacuum runs.
I have a good performance in my hard disks, I have a good amount of
memory, but my cores are very poor, only 1ghz each one.
I have some questions here:
1. autovacuum_max_workers= 3 , each work processes is using only one
core or one core it's sharing por 3 workers?
2. when I run a explain analyze in a very big table (30millons of
rows) , explain returning me 32 millons of rows moved, I am assuming
that my statistics are not updated in 2 millons of rows, but, is it a
very important number? or maybe, it's a regular result.
thanks for your help?
--
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance