On Sat, Jun 6, 2020 at 11:24 PM Wenjun Che <wenjun@openfin.co> wrote:
Hi
I am testing full vacuum with pg 10.10 on AWS RDS. I noticed for some tables, the number of waste bytes stays at a few MB after I run full vacuum. I double-checked that there are no long running transactions, no orphaned prepared transactions and no abandoned replication slots.
Here is output from full vacuum for one of the tables:
VACUUM(FULL, ANALYZE, VERBOSE) app_events_users vacuuming "app_events_users" "app_events_users": found 0 removable, 1198881 nonremovable row versions in 13369 pages analyzing "licensing.app_events_users" "app_events_users": scanned 13369 of 13369 pages, containing 1198881 live rows and 0 dead rows; 30000 rows in sample, 1198881 estimated total rows
What else can prevent full vacuum from reclaiming all waste space ?
Thank you
What "waste query" are you running? Those tend to be estimates only. Vacuum Full clearly did its job from that log you shared.