Обсуждение: postgres writer process growing up too much

Поиск
Список
Период
Сортировка

postgres writer process growing up too much

От
"Heiner Vega"
Дата:
Hi to everyone

I've been monitoring my postgres processes and I noticed that the resident memory
size of the writer process is growing up too much.

Those are reports from top in a 24 hour interval:

YESTERDAY:

 PID   USER      PR  NI  VIRT  RES  SHR  S %CPU %MEM    TIME+  COMMAND
11706 postgres   17   0  157m  148m 146m S  0.0   3.9   3:23.41 postgres: postgres BJ3 127.0.0.1(64648) idle
12174 postgres   16   0  157m  148m 146m S  0.0  3.9   3: 17.58 postgres: postgres BJ3 127.0.0.1(65367) idle
11495 postgres   16   0  157m  148m 146m S  5.7  3.9   7:51.98 postgres: postgres BJ3 127.0.0.1(64603) idle
11908 postgres   16   0  156m  147m 146m S  5.7  3.9   7:35.84 postgres: postgres BJ3 127.0.0.1 (64947) idle
11419 postgres   15   0  155m  112m 112m S  0.0  3.0   0:00.49 postgres: writer process

TODAY:
 PID   USER      PR  NI  VIRT  RES  SHR  S %CPU %MEM    TIME+  COMMAND
11706 postgres   17   0  157m  148m 146m S  0.0  3.9   7:37.16 postgres: postgres BJ3 127.0.0.1(64648) idle
12174 postgres   17   0  157m  148m 146m S  0.0  3.9   7:40.10 postgres: postgres BJ3 127.0.0.1(65367) idle
11495 postgres   16   0  157m  148m 146m S  6.3  3.9  17:46.99 postgres: postgres BJ3 127.0.0.1(64603) idle
11908 postgres   16   0  156m  147m 146m S  5.7  3.9  17:29.31 postgres: postgres BJ3 127.0.0.1(64947) idle
11419 postgres   15   0  155m  140m 139m S  0.0  3.7   0: 01.09 postgres: writer process

Any ideas why is this happening?
My postgres version is 8.1.3

Thanks a lot
--
Heiner Vega Thames

Re: postgres writer process growing up too much

От
Richard Huxton
Дата:
Heiner Vega wrote:
> Hi to everyone
>
> I've been monitoring my postgres processes and I noticed that the resident
> memory
> size of the writer process is growing up too much.

> YESTERDAY:
>
>  PID   USER      PR  NI  VIRT  RES  SHR  S %CPU %MEM    TIME+  COMMAND

> 11419 postgres   15   0  155m  112m 112m S  0.0  3.0   0:00.49 postgres:
> writer process
>
> TODAY:
>  PID   USER      PR  NI  VIRT  RES  SHR  S %CPU %MEM    TIME+  COMMAND

> 11419 postgres   15   0  155m  140m 139m S  0.0  3.7   0:01.09 postgres:
> writer process

Notice the "SHR"=shared value. That's 155MB virtual memory, 140MB of it
resident of which 139MB is shared with other processes. So - nothing to
worry about.

--
   Richard Huxton
   Archonet Ltd

Re: postgres writer process growing up too much

От
Tom Lane
Дата:
Richard Huxton <dev@archonet.com> writes:
> Heiner Vega wrote:
>> I've been monitoring my postgres processes and I noticed that the resident
>> memory
>> size of the writer process is growing up too much.

> Notice the "SHR"=shared value. That's 155MB virtual memory, 140MB of it
> resident of which 139MB is shared with other processes. So - nothing to
> worry about.

The reason the SHR number grows over time is that the system only counts
a page of shared memory against the process after the process has first
touched it.  Once the bgwriter has touched every page of shared buffers,
the number will stop changing.

If there were actually a memory leak, the VIRT number would be growing
... but it's not.

>> My postgres version is 8.1.3

If I were you, I'd be considerably more worried about the fact that
you are running an old release with a pile of known bugs.

            regards, tom lane

Re: postgres writer process growing up too much

От
"Heiner Vega"
Дата:
Thanks to you guys for your help... I appreciate it a lot.

Now, I still have my SHR and RES growing up. How can I know the number at which those values should stop?

On Dec 14, 2007 5:12 PM, Tom Lane < tgl@sss.pgh.pa.us> wrote:
Richard Huxton < dev@archonet.com> writes:
> Heiner Vega wrote:
>> I've been monitoring my postgres processes and I noticed that the resident
>> memory
>> size of the writer process is growing up too much.

> Notice the "SHR"=shared value. That's 155MB virtual memory, 140MB of it
> resident of which 139MB is shared with other processes. So - nothing to
> worry about.

The reason the SHR number grows over time is that the system only counts
a page of shared memory against the process after the process has first
touched it.  Once the bgwriter has touched every page of shared buffers,
the number will stop changing.

If there were actually a memory leak, the VIRT number would be growing
... but it's not.

>> My postgres version is 8.1.3

If I were you, I'd be considerably more worried about the fact that
you are running an old release with a pile of known bugs.

                       regards, tom lane



--
Heiner Vega Thames