| From: | Kevin Grittner <kgrittn(at)ymail(dot)com> |
|---|---|
| To: | Jeison Bedoya <jeisonb(at)audifarma(dot)com(dot)co>, "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org> |
| Subject: | Re: Process in state BIND, authentication, PARSE |
| Date: | 2013-07-10 13:37:33 |
| Message-ID: | 1373463453.36614.YahooMailNeo@web162904.mail.bf1.yahoo.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
Jeison Bedoya <jeisonb(at)audifarma(dot)com(dot)co> wrote:
> Ram: 128GB
> max_connections = 900
> temp_buffers = 128MB
Besides the concerns already expressed about work_mem, temp_buffers
could be a big problem. If a connection uses temp tables it
acquires up to 128MB, *and holds on it reserved for caching temp
tables for that connection for as long as the connection lasts*.
So, for 900 connections, that could be 112.5 GB. I would expect to
see performance decrease and eventually completely tank as more
connections reserved memory for this purpose.
--
Kevin Grittner
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Josh Berkus | 2013-07-10 16:47:34 | Re: [PERFORM] In progress INSERT wrecks plans on table |
| Previous Message | Rick Otten | 2013-07-10 13:12:16 | Re: Performance autovaccum |