Re: [GENERAL] Analyzing performance regression from 9.2 to 9.6

2017-09-09 Thread Dave Peticolas
> Am I right in guessing that report_submission.id is a > declared-not-null column, so that the join FROM "report_skilltype" > should be understood as an anti-join? You are exactly right. > Any chance of whacking your ORM upside the head to the point where it would emit that? In this case it

Re: [GENERAL] Analyzing performance regression from 9.2 to 9.6

2017-09-09 Thread Tom Lane
Dave Peticolas writes: > Hi, I am trying to analyze a performance regression from 9.2.21 to 9.6.3. > The query and execution plans are below with 9.6.3 first. Hm. Neither version is exactly covering itself with glory. I'm not sure why 9.6 doesn't pick the same plan as 9.2, but

Re: [GENERAL] pg_ident mapping Kerberos Usernames

2017-09-09 Thread rob stone
On Sat, 2017-09-09 at 20:44 -0500, techmail+pg...@dangertoaster.com wrote: > Hi, > > I'm trying to get pg_ident to map "user1" and "us...@a.domain.tld" > to > "user1" in postgres, or vice versa. I'm not picky about which way > works. > > Kerberos authentication works. I've gotten "user1" to

[GENERAL] pg_ident mapping Kerberos Usernames

2017-09-09 Thread techmail+pgsql
Hi, I'm trying to get pg_ident to map "user1" and "us...@a.domain.tld" to "user1" in postgres, or vice versa. I'm not picky about which way works. Kerberos authentication works. I've gotten "user1" to login successfully with a Kerberos ticket, but I'm not able to get "us...@a.domain.tld" to

[GENERAL] Analyzing performance regression from 9.2 to 9.6

2017-09-09 Thread Dave Peticolas
Hi, I am trying to analyze a performance regression from 9.2.21 to 9.6.3. The query and execution plans are below with 9.6.3 first. If the query looks a bit odd, it was generated by an ORM and the names have been modified. The slight difference in row counts is because they are from snapshots