Re: [GENERAL] Queries are taking way longer in 9.6 than 9.5
От | Chris Mair |
---|---|
Тема | Re: [GENERAL] Queries are taking way longer in 9.6 than 9.5 |
Дата | |
Msg-id | dfa066c765ed68b8564a30322467e6dc@smtp.hushmail.com обсуждение исходный текст |
Ответ на | Re: [GENERAL] Queries are taking way longer in 9.6 than 9.5 (Tomas Vondra <tomas.vondra@2ndquadrant.com>) |
Ответы |
Re: [GENERAL] Queries are taking way longer in 9.6 than 9.5
|
Список | pgsql-general |
On 17/01/17 23:21, Tomas Vondra wrote: > Hi, > > after looking at the explain plans again, I very much doubt those come from the same query. The reason is the 9.5 plancontains this part: > > -> HashAggregate (cost=67.54..68.12 rows=192 width=4) > Group Key: d.vip_patient_id > -> Nested Loop (cost=0.17..67.44 rows=192 width=4) > -> Index Scan using unq_user_name on tblcnaccounts ... > Index Cond: ((user_name)::text = 'dddddd'::text) > -> Index Only Scan using idx_tblcndoctorpatientmap ... > Index Cond: (master_user_id = a.master_user_id) > > while the 9.6 plan does not include anything like that, i.e. there's only a single aggregate at the top level, withoutany group keys. Also, the SQL query you've provided does not include any GROUP BY clause, so I claim that those plansare from two different queries. There's also a user_name = 'dddddd' that has become a user_name = 'rdoyleda' ... Ravi, could you please send the current query you're testing and the explain analyze of that query on 9.5 and 9.6? Bye, Chris.
В списке pgsql-general по дате отправления: