Anderson Valadares wrote:
[this is on Windows, DB is accessed with ODBC driver 8.4.3]
> Thanks for the answer ...
>
> But honestly I think that was a misunderstood.
>
>
>
> The memory increase issue is showed in the DATA column.
>
> Look how day by day it increases exponencially.
>
> In a few days PostGres goes out of memory, close the
> connections and enter in a recovery mode.
>
> I really don’t know what is causing it.
>
>
>
>
> Date 29/06/2009
>
> PID USER PR NI VIRT SWAP RES SHR DATA CODE S %CPU %MEM TIME+ COMMAND
> 9943 postgres 15 0 860m 41m 819m 811m 9604 3540 D 88.3 20.4 0:08.33 postgres: dbtest test 10.255.100.73(4796)
SELECT
>
> Date 29/06/2009
>
> PID USER PR NI VIRT SWAP RES SHR DATA CODE S %CPU %MEM TIME+ COMMAND
> 9943 postgres 15 0 994m 33m 960m 818m 143m 3540 S 29.5 23.9 48:19.96 postgres: dbtest test 10.255.100.73(4796)
idle
>
> Date 29/06/2009
>
> PID USER PR NI VIRT RES SHR CODE DATA S %CPU %MEM TIME+ COMMAND
> 9943 postgres 16 0 1366m 1.3g 818m 3540 515m S 31.2 33.2 192:20.61 postgres: dbtest test 10.255.100.73(4796)
SELECT
>
> Date 30/06/2009
>
> PID USER PR NI VIRT SWAP RES SHR DATA CODE S %CPU %MEM TIME+ COMMAND
> 9943 postgres 17 0 1724m 30m 1.7g 821m 873m 3540 R 27.2 42.2 325:54.83 postgres: citgis citgis
10.255.100.73(4796)SELECT
Now that is weird.
How can the same backend process suddenly be connected to database "citgis" as user "citgis"?
Do you have an explanation?
What is your work_mem setting?
This influences the amount of "private" memory a backend will allocate.
Can you say more that "executes a PL/pgSQL function in a loop" about the workload?
Are there long transactions?
Which version of PostgreSQL is this?
Yours,
Laurenz Albe