Re: [PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-24 Thread Tom Lane
Carmen Mardiros writes: > I've then tried different settings for work_mem, not changing anything else. > work_mem = 400MB -> query runs fine but memory usage in the system doesn't > exceed 1.3% > work_mem = 500MB -> usage hits 100% and postgres crashes out of memory. I suspect what may be happeni

Re: [PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-24 Thread Carmen Mardiros
It's PostgreSQL 9.5.4 On Thu, 24 Nov 2016 at 11:46 Imre Samu wrote: > > I am trying to configure Postgres (version 9.5) > > This is the latest PG 9.5 ? ( = 9.5.5 > ? [ > Release Date : 2016-10-27 ] ) > > 2016-11-24 9:17 GMT+01:

Re: [PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-24 Thread Imre Samu
> I am trying to configure Postgres (version 9.5) This is the latest PG 9.5 ? ( = 9.5.5 ? [ Release Date : 2016-10-27 ] ) 2016-11-24 9:17 GMT+01:00 Carmen Mardiros : > Hi Andreas, > > Thanks for the reply. fsync is off because

Re: [PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-24 Thread Carmen Mardiros
Hi Andreas, Thanks for the reply. fsync is off because I don't care if the data gets corrupted in this environment if it means a performance gain, as it's only a minor inconvenience to restore from backups. Here is the query itself that pushes memory usage to 100%: http://pastebin.com/VzCAerwd .

Re: [PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-23 Thread Andreas Kretschmer
Am 23. November 2016 23:15:25 MEZ, schrieb Carmen Mardiros : > >various combinations)? How can I investigate what's limiting postgres >from >doing so? Why fsync=off? Please run the queries with EXPLAIN ANALYSE and show us the output. -- Diese Nachricht wurde von meinem Android-Mobiltelefon mi

[PERFORM] How to tune Postgres to take advantage of 256GB RAM hardware

2016-11-23 Thread Carmen Mardiros
I am trying to configure Postgres (version 9.5) to take advantage of very large memory environment. Server configuration has 256GB RAM, 12 cores and 2 SSDs on RAID0 and runs Ubuntu. Swap is set at 4GB. The machine is used for data warehouse operations. Typically only 1 statement runs at a time (bu