Re: [BUGS] Giant TOAST tables due to many almost empty pages

2010-05-14 Thread Tom Lane
Rumko rum...@gmail.com writes: On Thursday 13. of May 2010 21:43:37 Tom Lane wrote: Do *any* of the rows in pg_class have non-null reloptions? First of all, really sorry. select reloptions from pg_class where relname = 'pg_toast_1066371'; Returns {autovacuum_enabled=false} (a remnant of some

Re: [BUGS] Giant TOAST tables due to many almost empty pages

2010-05-14 Thread Tom Lane
Rumko rum...@gmail.com writes: On Friday 14. of May 2010 19:29:44 Tom Lane wrote: Hmm, do both of the toast tables with bloat problems have {autovacuum_enabled=false} ? Yeah, but also many others that don't have the problem. Hmm, well I can reproduce the problem after doing alter table foo

Re: [BUGS] Giant TOAST tables due to many almost empty pages

2010-05-13 Thread Rumko
On Thursday 13. of May 2010 17:24:47 Tom Lane wrote: Rumko rum...@gmail.com writes: Tom Lane wrote: There's something extremely wacko about that vacuum output. Regarding storage paramaters, you mean ALTER TABLE x SET STORAGE...? Then no. No, I was wondering about ALTER TABLE ... SET

Re: [BUGS] Giant TOAST tables due to many almost empty pages

2010-05-12 Thread Tom Lane
Rumko rum...@gmail.com writes: INFO: vacuuming pg_toast.pg_toast_1066371 INFO: pg_toast_1066371: found 0 removable, 3259181 nonremovable row versions in 3259181 pages DETAIL: 0 dead row versions cannot be removed yet. Nonremovable row versions range from 57 to 122 bytes long. There were

Re: [BUGS] Giant TOAST tables due to many almost empty pages

2010-05-12 Thread Rumko
Tom Lane wrote: snip There's something extremely wacko about that vacuum output. A toast table should have few, if any, rows that short. And it's impossible to believe there's no free space at all in the table, especially since 122*3259181 bytes is still quite a lot less than 3259181