Обсуждение: postmaster stop hangs

Greetings,

Running /etc/init.d/postgresql stop results in:

    Stopping postgresql service:               [FAILED]

I ran an strace on "/usr/local/pgsql/bin/pg_ctl stop -D
/data/pgsql/data -s -m fast" and it is sending a SIGINT to processid 5676

Here is a ps of the currently running instance:

postgres  5671     1  0 02:32 pts/0    00:00:00
/usr/local/pgsql/bin/postmaster -D /data/pgsql/data
postgres  5674  5671  0 02:32 pts/0    00:00:00 postgres: writer process
postgres  5675  5671  0 02:32 pts/0    00:00:00 postgres: stats buffer
process
postgres  5676  5675  0 02:32 pts/0    00:00:00 postgres: stats collector
process

In this case, shouldn't it be sending a SIGTERM to 5671?

If I manually send a SIGINT or a SIGTERM to the postmaster (5671 in this
case), postgres obediently shuts down.

5676 is in the */data/postmaster.pid

Thanks for any help,
Drew.

postgresql 8.0.2
Linux xxx 2.6.9-5.0.3.EL #1 Sat Feb 19 18:26:49 CST 2005 i686 i686 i386
GNU/Linux

"Simpson" <asimpson@i-55.com> writes:
> Here is a ps of the currently running instance:

> postgres  5671     1  0 02:32 pts/0    00:00:00
> /usr/local/pgsql/bin/postmaster -D /data/pgsql/data
> postgres  5674  5671  0 02:32 pts/0    00:00:00 postgres: writer process
> postgres  5675  5671  0 02:32 pts/0    00:00:00 postgres: stats buffer
> process
> postgres  5676  5675  0 02:32 pts/0    00:00:00 postgres: stats collector
> process

> 5676 is in the */data/postmaster.pid

That seems extremely broken.  Can you reproduce this state (wrong PID in
postmaster.pid) after restarting the postmaster?

            regards, tom lane

Yes, it does it consistently.
Maybe I should start from scratch, if this seems very broken.
Drew.

-----Original Message-----
From: pgsql-admin-owner@postgresql.org
[mailto:pgsql-admin-owner@postgresql.org]On Behalf Of Tom Lane
Sent: Sunday, April 24, 2005 11:02 AM
To: Simpson
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] postmaster stop hangs


"Simpson" <asimpson@i-55.com> writes:
> Here is a ps of the currently running instance:

> postgres  5671     1  0 02:32 pts/0    00:00:00
> /usr/local/pgsql/bin/postmaster -D /data/pgsql/data
> postgres  5674  5671  0 02:32 pts/0    00:00:00 postgres: writer process
> postgres  5675  5671  0 02:32 pts/0    00:00:00 postgres: stats buffer
> process
> postgres  5676  5675  0 02:32 pts/0    00:00:00 postgres: stats collector
> process

> 5676 is in the */data/postmaster.pid

That seems extremely broken.  Can you reproduce this state (wrong PID in
postmaster.pid) after restarting the postmaster?

            regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
      joining column's datatypes do not match

"Simpson" <asimpson@i-55.com> writes:
> Yes, it does it consistently.
> Maybe I should start from scratch, if this seems very broken.

It seems not only very broken, but practically impossible: there is only
one routine that writes that file, and it definitely writes the result
of getpid(), and it is definitely not called in the stats collector
process.  How did you obtain or build your copy of PG?  What is the
platform exactly (you said Linux, but what release of what distro)?

            regards, tom lane