Re: fsm and vacuum
От | Guillaume Lelarge |
---|---|
Тема | Re: fsm and vacuum |
Дата | |
Msg-id | 4CF8158A.5080208@lelarge.info обсуждение исходный текст |
Ответ на | fsm and vacuum ("Little, Douglas" <DOUGLAS.LITTLE@orbitz.com>) |
Список | pgadmin-support |
Le 02/12/2010 22:00, Little, Douglas a écrit : > We're new to Greenplum, based on PG (8.2.13). we weren't advised to increase the max_fsm_relations switch as our db hasgrown. > Currently we're nearly 14k tables/indexes and the switch is set to 1000. > We've got it updated now, but wondering about the effect & recovery. > > We've regularly been reloading tables with CTAS. > I've done some experiments and want to confirm my understanding. > > 1. Vacuum full only compacted the oldest? tables (reducing pages from 109 to 5) > 2. vacuum full frees pages, but doesn't reduce the file size (newer tables). Table needed to be recreated to reduce size > > I've searched the archives and haven't found what I'm looking for. I hope someone would just dash off a simple explanation. > > 1. when are relations added to the fsm? (at create, or some other time) During vacuum, if there are dead spaces in the table. > 2. will vacuum (full) add a relation to the fsm - ? Yes, for tables with dead spaces. > 3. if bloated table is vacuum full'd, will the free space pages be use? If vacuum full did its job, it shouldn't be. > My assumptions are: > 1. relation added to fsm list at create time. Nope. > 2. fsm_pages keep track of free space for the relations being tracked in the fsm_relation list At vacuum time. > 3. table recreate is the only way to have a relation tracked. Nope. -- Guillaumehttp://www.postgresql.frhttp://dalibo.com
В списке pgadmin-support по дате отправления: