Обсуждение: postmaster respawn....?

Поиск
Список
Период
Сортировка

postmaster respawn....?

От
jerome
Дата:
im sure there is only one postmaster running on my server but everytime i do
a series of ps ax | grep post..

**1st ps

  510 ?        R      4:25 /usr/bin/postmaster -o -F -i
 1235 ?        S    119:12 /usr/bin/postgres localhost nsadmin openacs idle
 1394 ?        S     74:00 /usr/bin/postgres localhost nsadmin openacs idle
 1422 ?        S      1:39 /usr/bin/postgres localhost nsadmin openacs idle
26375 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
30626 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
30632 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
32238 ?        R      0:26 /usr/bin/postgres 127.0.0.1 nsadmin openacs SELECT
  443 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
  449 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle

**2nd ps
  510 ?        R      4:26 /usr/bin/postmaster -o -F -i
 1235 ?        S    119:15 /usr/bin/postgres localhost nsadmin openacs idle
 1394 ?        S     74:00 /usr/bin/postgres localhost nsadmin openacs idle
 1422 ?        S      1:40 /usr/bin/postgres localhost nsadmin openacs idle
30626 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
30632 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
  443 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
  449 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
 1971 ?        R      0:00 /usr/bin/postmaster -o -F -i

**3rd ps
  510 ?        R      4:25 /usr/bin/postmaster -o -F -i
 1235 ?        S    119:12 /usr/bin/postgres localhost nsadmin openacs idle
 1394 ?        S     74:00 /usr/bin/postgres localhost nsadmin openacs idle
 1422 ?        S      1:39 /usr/bin/postgres localhost nsadmin openacs idle
26375 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
30626 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
30632 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
32238 ?        R      0:30 /usr/bin/postgres 127.0.0.1 nsadmin openacs SELECT
  443 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
  449 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
  633 ?        Z      0:00 [postmaster <defunct>]


then it cycles again...  hmmm...  can anyone tell me whats happening? does
this hamper my servers efficiency?


thanks...

Re: postmaster respawn....?

От
Christoph Dalitz
Дата:
On Fri, 30 Aug 2002 09:18:07 -0400
pgsql-general-owner@postgresql.org wrote:
>
> I'm sure there is only one postmaster running on my server but everytime i do
> a series of ps ax | grep post..
>
> **1st ps
>
>   510 ?        R      4:25 /usr/bin/postmaster -o -F -i
>  1235 ?        S    119:12 /usr/bin/postgres localhost nsadmin openacs idle
>
That is perfectly sane: for each db connection or session a new backend is
spawned. It is killed when the connection is closed.

Thus you see on postmaster (= PG daemon) and several backends.

Christoph Dalitz

Re: postmaster respawn....?

От
Tom Lane
Дата:
jerome <jerome@gmanmi.tv> writes:
> im sure there is only one postmaster running on my server but everytime i do
> a series of ps ax | grep post..

> **2nd ps
>   510 ?        R      4:26 /usr/bin/postmaster -o -F -i
>  1235 ?        S    119:15 /usr/bin/postgres localhost nsadmin openacs idle
>  1394 ?        S     74:00 /usr/bin/postgres localhost nsadmin openacs idle
>  1422 ?        S      1:40 /usr/bin/postgres localhost nsadmin openacs idle
> 30626 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
> 30632 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
>   443 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
>   449 ?        S      0:00 /usr/bin/postgres localhost nsadmin openacs idle
>  1971 ?        R      0:00 /usr/bin/postmaster -o -F -i

It would be more clear what was happening if you used ps flags that
would include the parent process ID in the listing.  My guess is that
process 1971 is a newly-forked backend that hasn't had a chance to
change its ps display yet.  However, the window for that is pretty durn
short, so it's surprising that you'd see this on any regular basis.

If 1971 is not a child of 510 but of someone else, then you've got
trouble --- but finding out who the someone else is would be the first
step towards understanding the issue anyway.

            regards, tom lane