Re: BUG #8318: memory leak during CREATE TEMPORARY TABLE

"G.Bakalarski@icm.edu.pl" <G.Bakalarski@icm.edu.pl> wrote:=0A> Kevin Grittn=
er <kgrittn@ymail.com> wrote:=0A>> "g.bakalarski@icm.edu.pl" <g.bakalarski@=
icm.edu.pl> wrote:=0A>>=0A>>> When doing these kind of statements=0A>>=0A>>=
> execute <unnamed>: CREATE TEMPORARY TABLE [...]=0A>>=0A>>> After few days=
 [...] my huge server (192GB of RAM), gets stucked=0A>>> due to no memory a=
vailable (usually %commit is around 103-105%),=0A>>> linux OOM killer goes =
into action - it usually kills one of=0A>>> postgres processes - postmaster=
 restart all databases) and after=0A>>> all linux=A0 has again some 100GB o=
f not commited memory=0A>>=0A>> What do you have as settings for temp_buffe=
rs and max_connections?=0A>>=0A> temp_buffers =3D 16GB=0A> max_connections =
=3D 250=0A=0Atemp_buffers is per-connection, and any space allocated for th=
is=0Apurpose by a connection is not released until the connection is=0Aclos=
ed.=A0 So you have configured your server to acquire and hold up=0Ato 3.9 T=
B of RAM for this purpose.=0A=0Ahttp://www.postgresql.org/docs/current/stat=
ic/runtime-config-resource.html#RUNTIME-CONFIG-RESOURCE-MEMORY=0A=0A> (with=
 real number of connections about 120 and not going above)=0A=0AOK, so you =
can only get to 1.9 TB of RAM used for temp_buffers with=0A120 connections,=
 but that's still way more than the 192 GB you=0Aactually have.=0A=0A--=0AK=
evin Grittner=0AEDB: http://www.enterprisedb.com=0AThe Enterprise PostgreSQ=
L Company