Обсуждение: pg_ctlcluster is not stopping cluster

On 4/7/23 15:27, Telium Technical Support wrote:
> I am string to stop my PostgreSQL (on debian 11) server using the 
> following command
> 
> root@d11:/# sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -- -m 
> fast -D /var/lib/postgresql/13/main
> 
> Notice: extra pg_ctl/postgres options given, bypassing systemctl for 
> stop operation
> 
> pg_ctl: PID file "/var/lib/postgresql/13/main/postmaster.pid" does not exist
> 
> Is server running?
> 
> The notice is correct, the is no such postmaster.pid file in that 
> directory, but yes the service is running. I can confirm it's running with:
> 
> root@d11:/# sudo -u postgres /usr/bin/pg_ctlcluster 15 main status -- -D 
> /var/lib/postgresql/13/main
> 
> pg_ctl: server is running (PID: 2701882)
> 
> /usr/lib/postgresql/15/bin/postgres "-D" "/var/lib/postgresql/13/main" 
> "-c" "config_file=/etc/postgresql/15/main/postgresql.conf"
> 
> So why is my stop command being ignored? (I tried without the -m fast 
> option but no change, and I'd like to keep that for other reasons). I 
> confirmed with 'ps ax' that postgresql 15 is running, despite the 
> directory suggesting it might be 13.  Coincidentally, there is a 
> postmaster.pid file in a directory OTHER than the data directory:
> 
> /var/lib/postgresql/15/main/postmaster.pid
> 
> (and notice the 15). Is this a clue?

Yes that this:

  sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -- -m fast -D 
/var/lib/postgresql/13/main

is not correct.

First do:

pg_lsclusters

to determine what is actually running.

Then do

sudo -u postgres /usr/bin/pg_ctlcluster <version> main stop -- -m fast 
-D /var/lib/postgresql/13/main stop -m fast

for whatever version is running.

> 

-- 
Adrian Klaver
adrian.klaver@aklaver.com

On Fri, 7 Apr 2023, Telium Technical Support wrote:

>Notice: extra pg_ctl/postgres options given, bypassing systemctl for stop

>it might be 13.  Coincidentally, there is a postmaster.pid file in a
>directory OTHER than the data directory:
>
>/var/lib/postgresql/15/main/postmaster.pid
>
>(and notice the 15). Is this a clue?

Maybe the pidfile is written into the cluster version-based directory
instead of the data directory. Best figure out what exactly writes the
pidfile, whether systemd is used for starting (which would of course be
a prime suspect as it says it’s explicitly not used for stopping), etc.

Maybe this is indeed a bug in whatever determines the pidfile path,
perhaps not; is it supposed to live within the data directory?

bye,
//mirabilos
--
15:41⎜<Lo-lan-do:#fusionforge> Somebody write a testsuite for helloworld :-)

On Fri, 7 Apr 2023, Telium Technical Support wrote:

>I’m assuming something is misconfigured on the host that’s causing this
>unusual behavior….and that’s what I need to understand

The mix between 13 and 15 here is what I’d consider a misconfiguration.

Also, please don’t top-post and full-quote.

bye,
//mirabilos
--
15:41⎜<Lo-lan-do:#fusionforge> Somebody write a testsuite for helloworld :-)

I tried the command you suggested, and it shows that data directory status as "DOWN".  Yet when I ask pg_ctlcluster for
it'sstatus it says the server is running. 

What does this mean?

root@d11:/var/tmp/myapp# pg_lsclusters
Ver Cluster Port Status Owner    Data directory              Log file
13  main    5432 down   postgres /var/lib/postgresql/13/main /var/log/postgresql/postgresql-13-main.log
15  main    5433 down   postgres /var/lib/postgresql/15/main /var/log/postgresql/postgresql-15-main.log
root@d11:/var/tmp/myapp# sudo -u postgres /usr/bin/pg_ctlcluster 15 main status -- -D /var/lib/postgresql/13/main
pg_ctl: server is running (PID: 2701882)
/usr/lib/postgresql/15/bin/postgres "-D" "/var/lib/postgresql/13/main" "-c"
"config_file=/etc/postgresql/15/main/postgresql.conf"
root@d11:/var/tmp/myapp#

-----Original Message-----
From: Adrian Klaver [mailto:adrian.klaver@aklaver.com]
Sent: Friday, April 7, 2023 6:47 PM
To: Telium Technical Support <support@telium.io>; pgsql-general@lists.postgresql.org
Subject: Re: pg_ctlcluster is not stopping cluster

On 4/7/23 15:27, Telium Technical Support wrote:
> I am string to stop my PostgreSQL (on debian 11) server using the
> following command
>
> root@d11:/# sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -- -m
> fast -D /var/lib/postgresql/13/main
>
> Notice: extra pg_ctl/postgres options given, bypassing systemctl for
> stop operation
>
> pg_ctl: PID file "/var/lib/postgresql/13/main/postmaster.pid" does not
> exist
>
> Is server running?
>
> The notice is correct, the is no such postmaster.pid file in that
> directory, but yes the service is running. I can confirm it's running with:
>
> root@d11:/# sudo -u postgres /usr/bin/pg_ctlcluster 15 main status --
> -D /var/lib/postgresql/13/main
>
> pg_ctl: server is running (PID: 2701882)
>
> /usr/lib/postgresql/15/bin/postgres "-D" "/var/lib/postgresql/13/main"
> "-c" "config_file=/etc/postgresql/15/main/postgresql.conf"
>
> So why is my stop command being ignored? (I tried without the -m fast
> option but no change, and I'd like to keep that for other reasons). I
> confirmed with 'ps ax' that postgresql 15 is running, despite the
> directory suggesting it might be 13.  Coincidentally, there is a
> postmaster.pid file in a directory OTHER than the data directory:
>
> /var/lib/postgresql/15/main/postmaster.pid
>
> (and notice the 15). Is this a clue?

Yes that this:

  sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -- -m fast -D /var/lib/postgresql/13/main

is not correct.

First do:

pg_lsclusters

to determine what is actually running.

Then do

sudo -u postgres /usr/bin/pg_ctlcluster <version> main stop -- -m fast -D /var/lib/postgresql/13/main stop -m fast

for whatever version is running.

>

--
Adrian Klaver
adrian.klaver@aklaver.com

>The mix between 13 and 15 here is what I’d consider a misconfiguration.

As I inherited this (and I'm somewhat new to pgsql), I'm trying to understand this.  From the docs I read online, the
postmaster.pidfile is supposed to reside in the data directory.  Which it does.  So that's ok. 

Does the fact that the database resides in the /var/lib/postgresql/13 mean I have multiple pgsql servers running?
(Doesthe directory name make a big difference)  Can I just kill all pgsql processes and move the directory into the /15
directoryand problem solved? 

It feels like I'm missing something obvious...why would the directory matter so much (since pgsql is clearly tracking
itin the right dir) 

On 4/7/23 15:51, Telium Technical Support wrote:
> I tried the command you suggested, and it shows that data directory status as "DOWN".  Yet when I ask pg_ctlcluster
forit's status it says the server is running.
 
> 
> What does this mean?
> 
> root@d11:/var/tmp/myapp# pg_lsclusters
> Ver Cluster Port Status Owner    Data directory              Log file
> 13  main    5432 down   postgres /var/lib/postgresql/13/main /var/log/postgresql/postgresql-13-main.log
> 15  main    5433 down   postgres /var/lib/postgresql/15/main /var/log/postgresql/postgresql-15-main.log

Neither cluster is running.

To confirm do:

ps ax | grep postgres

> root@d11:/var/tmp/myapp# sudo -u postgres /usr/bin/pg_ctlcluster 15 main status -- -D /var/lib/postgresql/13/main
> pg_ctl: server is running (PID: 2701882)

Best guess, is that because of this:

sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -- -m fast -D 
/var/lib/postgresql/13/main

the Postgres pid did not get removed on shutdown.

Bottom line you should not use

pg_ctlcluster 15 main stop

to

shut down a 13 cluster located at:

-D /var/lib/postgresql/13/main

Just do:

sudo -u postgres /usr/bin/pg_ctlcluster 15 main stop -m fast

or

sudo -u postgres /usr/bin/pg_ctlcluster 13 main stop -- -m fast

depending on which cluster you want to shut down.


> /usr/lib/postgresql/15/bin/postgres "-D" "/var/lib/postgresql/13/main" "-c"
"config_file=/etc/postgresql/15/main/postgresql.conf"
> root@d11:/var/tmp/myapp#
> 

-- 
Adrian Klaver
adrian.klaver@aklaver.com

Thorsten Glaser <tg@evolvis.org> writes:

> On Fri, 7 Apr 2023, Telium Technical Support wrote:
>
>>Notice: extra pg_ctl/postgres options given, bypassing systemctl for stop
>
>>it might be 13.  Coincidentally, there is a postmaster.pid file in a
>>directory OTHER than the data directory:
>>
>>/var/lib/postgresql/15/main/postmaster.pid
>>
>>(and notice the 15). Is this a clue?
>
> Maybe the pidfile is written into the cluster version-based directory
> instead of the data directory. Best figure out what exactly writes the
> pidfile, whether systemd is used for starting (which would of course be



Some distros in their wrapper foo, I believe including Debian use this
setting to put a PID file under /var/run/postgresql IIRC but YMMV...

external_pid_file (string)

 Specifies the name of an additional process-ID (PID) file that the
 server should create for use by server administration programs. This
 parameter can only be set at server start.


> a prime suspect as it says it’s explicitly not used for stopping), etc.
>
> Maybe this is indeed a bug in whatever determines the pidfile path,
> perhaps not; is it supposed to live within the data directory?
>
> bye,
> //mirabilos