Обсуждение: PostgreSQL 8.3.1 on FreeBSD 6.0 - accumulation of processes

Hi all,

I am runnning PostgreSQL 8.3.1 on FreeBSD 6.0.
There are about 30 heavy readed / updated databases and very often is
occurred situation that there accumulate processes and waiting for each
other, for example:

when server starts, there are a few processes:


ps ax |grep postgres
50120  ??  Ss     0:03,61 /usr/local/bin/postmaster -D
/usr/local/pgsql/data -i (postgres)
50122  ??  S      0:01,11 postmaster: writer process    (postgres)
50123  ??  S      0:22,20 postmaster: stats buffer process    (postgres)
50125  ??  S      0:36,55 postmaster: stats collector process    (postgres)
53605  ??  IW     0:00,00 postmaster: postgres b2c_91355 [local] idle
(postgres)
59643  ??  D      0:01,35 postmaster: b2c_101394 b2c_101394 ::1(53787)
SELECT (postgres)
59663  ??  D      0:00,11 postmaster: b2c_90273 b2c_90273 ::1(60017)
SELECT (postgres)
59801  ??  S      0:00,85 postmaster: b2c_90273 b2c_90273 ::1(65358)
SELECT (postgres)
60076  ??  D      0:00,67 postmaster: b2c_101924 b2c_101924 ::1(53070)
SELECT (postgres)
60145  ??  S      0:00,11 postmaster: b2c_develop b2c_develop ::1(63134)
idle (postgres)
60220  ??  D      0:00,49 postmaster: b2c_103703 b2c_103703 ::1(49875)
SELECT (postgres)
60362  ??  D      0:00,83 postmaster: b2c_95002 b2c_95002 ::1(56467)
SELECT (postgres)
60631  ??  D      0:00,01 postmaster: b2c_95398 b2c_95398 ::1(64013)
SELECT (postgres)

but after a while processes dramaticaly grow up and there are about tens
  or hunderds processes.

Typical situation is, that ended sql commands not ends postmaster
processes. For example when I run vacuum full analyze some database, sql
command ends, but process not! Then I can see this:

62123  ??  D      0:22,24 postmaster: postgres b2c_100091 ::1(50512)
VACUUM (postgres)
64332  ??  I      0:00,05 postmaster: b2c_100091 b2c_100091 ::1(64940)
SELECT waiting (postgres)
64826  ??  S      0:00,06 postmaster: b2c_100091 b2c_100091 ::1(60703)
SELECT waiting (postgres)
64904  ??  S      0:00,06 postmaster: b2c_100091 b2c_100091 ::1(58878)
SELECT waiting (postgres)

Important is, that SQL command vacuum full analyze some database has
finished. I think, that when vacuum is finished, process

62123  ??  D      0:22,24 postmaster: postgres b2c_100091 ::1(50512)
VACUUM (postgres)

should end and dissapear.

Please, advice me where to look to solve this problem.

Ute

On Thu, Mar 16, 2006 at 05:17:47PM +0100, ute wrote:
> Typical situation is, that ended sql commands not ends postmaster
> processes. For example when I run vacuum full analyze some database, sql
> command ends, but process not! Then I can see this:
>
> 62123  ??  D      0:22,24 postmaster: postgres b2c_100091 ::1(50512)
> VACUUM (postgres)
> 64332  ??  I      0:00,05 postmaster: b2c_100091 b2c_100091 ::1(64940)
> SELECT waiting (postgres)
> 64826  ??  S      0:00,06 postmaster: b2c_100091 b2c_100091 ::1(60703)
> SELECT waiting (postgres)
> 64904  ??  S      0:00,06 postmaster: b2c_100091 b2c_100091 ::1(58878)
> SELECT waiting (postgres)
>
> Important is, that SQL command vacuum full analyze some database has
> finished. I think, that when vacuum is finished, process
>
> 62123  ??  D      0:22,24 postmaster: postgres b2c_100091 ::1(50512)
> VACUUM (postgres)

And do you have reason to believe that process isn't ending when vacuum
ends? Also, are you sure you need to be running vacuum full? It blocks
access to each table as it runs, which will undoubtedly cause processes
to back up waiting for access (which is exactly what that output of ps
shows).

My advice is to just let autovacuum handle things if at all possible.
--
Jim C. Nasby, Sr. Engineering Consultant      jnasby@pervasive.com
Pervasive Software      http://pervasive.com    work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf       cell: 512-569-9461