Re: Sudden drop in DBb performance
От | Gerhard Wohlgenannt |
---|---|
Тема | Re: Sudden drop in DBb performance |
Дата | |
Msg-id | 4E65DC88.3020206@ai.wu.ac.at обсуждение исходный текст |
Ответ на | Re: Sudden drop in DBb performance (Andy Colson <andy@squeakycode.net>) |
Список | pgsql-performance |
> That's a good thought, maybe the stats are old and you have bad > plans? It could also be major updates to the data too (as opposed to > growth). we have made checks for number of dead tuples etc recently, but looks ok. and as "everything" in the database seems to be very slow atm, I guess the problem is not caused by bad plans for specific tables/queries. > > Gerhard, have you done an 'explain analyze' on any of your slow > queries? Have you done an analyze lately? > yes we added the 'auto_explain' module to log/analyze queries >= 5000ms. a sample result from the logs (there is lots of stuff in the logs, I selected this query because it is very simple): 2011-09-06 04:00:35 CEST ANWEISUNG: INSERT into keywords.table_x_site_impact (content_id, site_impact_id, site_impact) VALUES (199083087, 1, 1.000000) 2011-09-06 04:00:35 CEST LOG: Dauer: 15159.723 ms Anweisung: INSERT into keywords.table_x_site_impact (content_id, site_impact_id, site_impact) VALUES (199083087, 1 , 1.000000) 2011-09-06 04:00:35 CEST LOG: duration: 15159.161 ms plan: Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.017..0.019 rows=1 loops=1) Output: nextval('keywords.table_x_site_impact_internal_id_seq'::regclass), 199083087::bigint, 1::smallint, 1::double precision
В списке pgsql-performance по дате отправления: