Skip to main content

Timeline for answer to Optimise PostgreSQL for fast testing by Craig Ringer

Current License: CC BY-SA 3.0

Post Revisions

40 events
when toggle format what by license comment
Oct 23, 2018 at 0:57 comment added Craig Ringer @DylanYoung Er... yeah. That.
Oct 22, 2018 at 18:40 comment added DylanYoung "For production use you can possibly use synchronous_commit=off and set a commit_delay" <-- commit_delay doesn't do anything when sync commit is off. Should that "and" be an "or"? ;)
Jun 2, 2017 at 0:10 comment added John Smith Can you maybe help me with this question as well: stackoverflow.com/questions/44318816/…
May 23, 2017 at 11:54 history edited URL Rewriter Bot
replaced http://stackoverflow.com/ with https://stackoverflow.com/
Jan 13, 2017 at 11:08 comment added Sajeev See this also: tekadempiere.blogspot.ae/2014/09/…
Dec 18, 2016 at 5:54 comment added Craig Ringer @maniek It should be, but I'd avoid relying on it for anything really important, just to be sure.
Dec 16, 2016 at 12:46 comment added maniek Is fsync=off safe against unclean postgres restarts (as opposed to OS restarts)?
Jul 20, 2016 at 13:40 comment added Jonathan Crosmer In fact, DELETE was much faster than TRUNCATE for our case
Jul 20, 2016 at 13:32 comment added Jonathan Crosmer stackoverflow.com/questions/11419536/… DELETE may be faster than TRUNCATE for tables with few rows, which is likely to be the case in tests.
Jun 30, 2015 at 11:53 comment added brauliobo @CraigRinger yeah, it is automated testing on rake tasks (rails)
Jun 30, 2015 at 11:38 comment added Craig Ringer @brauliobo My tests often do many tx's at high TPS ... because I try to simulate production, including concurrency-heavy workloads. If you mean "single-connection, linear testing" then I'd agree with you.
Jun 30, 2015 at 11:30 comment added brauliobo @CraigRinger it is useful in production, but not for testing
Jun 30, 2015 at 0:37 comment added Craig Ringer @brauliobo Why isn't that useful? It's enough to group up concurrent sets of commits, and that's what it's for.
Jun 29, 2015 at 21:41 comment added brauliobo unfortunetely commit_delay only allow a maximum of 100000 (100ms), so it is not useful...
Jun 3, 2015 at 14:21 history edited Craig Ringer CC BY-SA 3.0
added 13 characters in body
Jun 16, 2014 at 0:55 history edited Craig Ringer CC BY-SA 3.0
added 130 characters in body
Mar 26, 2014 at 3:15 history edited Craig Ringer CC BY-SA 3.0
added 673 characters in body
Nov 24, 2013 at 1:20 history edited Craig Ringer CC BY-SA 3.0
deleted 162 characters in body
Sep 19, 2013 at 14:18 comment added Gunnlaugur Briem Great writeup. Just as a tiny update, “You may need to increase the OS's maximum shared memory limit if you increase shared_buffers” is no longer true (for most users) under PostgreSQL 9.3: postgresql.org/docs/9.3/static/release-9-3.html#AEN114343
Jul 19, 2013 at 3:24 history wiki removed Brad Larson
Jul 19, 2013 at 1:24 history edited Craig Ringer CC BY-SA 3.0
added 147 characters in body
Jul 16, 2013 at 2:38 history edited Craig Ringer CC BY-SA 3.0
added 322 characters in body
Mar 1, 2013 at 0:43 history edited Craig Ringer CC BY-SA 3.0
added 104 characters in body
Nov 20, 2012 at 0:00 comment added Craig Ringer @JustBob Also, after a few months after release the issues are generally largely sorted out in the new major release. There are a few signs there might be a couple of nasty planner regressions in 9.2, but the trouble is that if nobody tests it seriously with their apps then issues aren't found. I won't jump straight to a new major in production, but I'll start testing it very promptly.
Nov 19, 2012 at 23:57 comment added Craig Ringer @JustBob No, it isn't, but I find it wise to stay within a major version or two. I'm really aiming at those people still on 8.3 or older, which is just absurd. 8.4 is marginal, I'd want to be on 9.1 at the moment and using 9.2 in my testing in preparation for rolling it out in the next six months or so. You save so much time using newer versions because issues you're facing are already solved in the new version; witness the vismap and fsm, autovac improvements, etc etc.
Nov 19, 2012 at 14:18 comment added Kuberchaun While I totally agree the newer version always have the best, they may also contain the worst(some new issue\bug). Is it really sane to try and stay lock step with the Postgres release cycle? For sure not in production. I'm honestly just curious of your input on this not trying to poke a stick at the lion.
Nov 8, 2012 at 1:49 history edited Craig Ringer CC BY-SA 3.0
added 1212 characters in body
Oct 17, 2012 at 12:51 history edited Craig Ringer CC BY-SA 3.0
added 490 characters in body
Aug 21, 2012 at 0:17 history edited Craig Ringer CC BY-SA 3.0
Some updates, formatting improvements and references; added 167 characters in body
Aug 17, 2012 at 4:37 history edited Craig Ringer CC BY-SA 3.0
Add link to http://www.postgresql.org/docs/9.1/static/non-durability.html
Aug 2, 2012 at 0:06 history edited Craig Ringer CC BY-SA 3.0
added 566 characters in body
Feb 27, 2012 at 0:27 vote accept Dmytrii Nagirniak
Feb 24, 2012 at 1:27 comment added Greg Smith I didn't release an update to the book for PostgreSQL 9.1, the only release since its publication, because there weren't enough performance related changes in 9.1 to warrant it.
Feb 23, 2012 at 16:22 comment added Daniel Lyons I love that book too. No, it's not an issue. The book covers up through 9.0 and PG is only up to 9.1.
Feb 23, 2012 at 9:26 comment added Dmytrii Nagirniak Thanks guys, will definitely put it on my reading queue. I suppose the fact the book is 2 years old isn't an issue since probably not much has changes in PG itself. Right?
Feb 23, 2012 at 7:59 comment added tscho I can also recommend PostgreSQL 9.0 High Performance by @GregSmith, it's really a great read. The book covers every aspect of performance tuning from disk layout to query tuning and gives you a very good understanding of the PG internals.
Feb 23, 2012 at 6:23 history edited Craig Ringer CC BY-SA 3.0
added 2208 characters in body
Feb 23, 2012 at 6:18 history edited Craig Ringer CC BY-SA 3.0
added 2208 characters in body
Feb 23, 2012 at 6:10 history edited Craig Ringer CC BY-SA 3.0
added 2208 characters in body
Feb 23, 2012 at 5:57 history answered Craig Ringer CC BY-SA 3.0