Обсуждение: it refuses to go down...
Hi Everybody,
I ran out of space over the weekend. /usr/local/ filled up.
I got a new partition and about 1/2 done with copying the
data from the filled-up partition. BTW, I am using postgres
8.3.0 on dell hardware/ redhat advanced server.
I did:
$ pg_ctl status
pg_ctl: server is running (PID: 11148)
/usr/local/pgsql/bin/postgres
and proceeded with:
$ pg_ctl stop -m fast
waiting for server to shut down............................................................... failed
pg_ctl: server does not shut down
My plan was to shutdown, make sure the env variables are
set correctly and restart. But it refuses to go down as
shown above. I then tried to use psql:
$ psql canon
and it tells me:
psql: FATAL: the database system is shutting down
Can I kill the process 11148 and do as I planned accordingly?
I would appreciate any pointers.
Regards,
Tena Sakai
tsakai@gallo.ucsf.edu
"Tena Sakai" <tsakai@gallo.ucsf.edu> writes:
> $ pg_ctl stop -m fast
> waiting for server to shut down............................................................... failed
> pg_ctl: server does not shut down
Have you looked into the postmaster log to see *why* it's not shutting
down?
> Can I kill the process 11148 and do as I planned accordingly?
It's not exactly recommended when you don't know what the problem is.
Killing only the postmaster and not its child processes is especially
not recommended --- there are some defenses in place against that,
but nothing that can't be broken by a sufficiently determined DBA.
regards, tom lane
That's what she said. Someone had to do it. That's what she said, again. Sorry. -salman
On Mon, Mar 24, 2008 at 3:48 PM, Tena Sakai <tsakai@gallo.ucsf.edu> wrote: > > Hi Everybody, > > I ran out of space over the weekend. /usr/local/ filled up. > I got a new partition and about 1/2 done with copying the > data from the filled-up partition. BTW, I am using postgres > 8.3.0 on dell hardware/ redhat advanced server. > > I did: > $ pg_ctl status > pg_ctl: server is running (PID: 11148) > /usr/local/pgsql/bin/postgres > > and proceeded with: > > $ pg_ctl stop -m fast > waiting for server to shut > down............................................................... failed > pg_ctl: server does not shut down did you try pg_ctl -m immediate stop ???
Hi Tom,
> Have you looked into the postmaster log to see
> *why* it's not shutting down?
I did tail on the log and got...
[2008-03-24 14:31:39.331 PDT] < 11148 2008-02-26 18:15:00 PST >LOG: received smart shutdown request
[2008-03-24 14:31:39.331 PDT] < 29578 2008-03-24 06:38:33 PDT >LOG: autovacuum launcher shutting down
[2008-03-24 14:32:58.740 PDT] < 11148 2008-02-26 18:15:00 PST >LOG: received fast shutdown request
[2008-03-24 14:46:03.123 PDT] <postgres 8619 127.0.0.1 2008-03-24 14:46:03 PDT /usr/local/pgsql/bin/postgres>FATAL: the database system is shutting down
[2008-03-24 15:52:43.520 PDT] <michaelstevens 11367 [local] 2008-03-24 15:52:43 PDT /usr/local/pgsql/bin/postgres>FATAL: the database system is shutting down
[2008-03-24 15:52:43.563 PDT] <michaelstevens 11368 [local] 2008-03-24 15:52:43 PDT /usr/local/pgsql/bin/postgres>FATAL: the database system is shutting down
I am about to try:
pg_ctl stop -m immediate
Regards,
Tena Sakai
-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Mon 3/24/2008 3:11 PM
To: Tena Sakai
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] it refuses to go down...
"Tena Sakai" <tsakai@gallo.ucsf.edu> writes:
> $ pg_ctl stop -m fast
> waiting for server to shut down............................................................... failed
> pg_ctl: server does not shut down
Have you looked into the postmaster log to see *why* it's not shutting
down?
> Can I kill the process 11148 and do as I planned accordingly?
It's not exactly recommended when you don't know what the problem is.
Killing only the postmaster and not its child processes is especially
not recommended --- there are some defenses in place against that,
but nothing that can't be broken by a sufficiently determined DBA.
regards, tom lane
Hi Scott,
> did you try pg_ctl -m immediate stop ???
I just did, and it worked.
$ pg_ctl stop -m immediate
waiting for server to shut down.... done
server stopped
Many thanks.
Regards,
Tena Sakai
tsakai@gallo.ucsf.edu
-----Original Message-----
From: Scott Marlowe [mailto:scott.marlowe@gmail.com]
Sent: Mon 3/24/2008 3:37 PM
To: Tena Sakai
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] it refuses to go down...
On Mon, Mar 24, 2008 at 3:48 PM, Tena Sakai <tsakai@gallo.ucsf.edu> wrote:
>
> Hi Everybody,
>
> I ran out of space over the weekend. /usr/local/ filled up.
> I got a new partition and about 1/2 done with copying the
> data from the filled-up partition. BTW, I am using postgres
> 8.3.0 on dell hardware/ redhat advanced server.
>
> I did:
> $ pg_ctl status
> pg_ctl: server is running (PID: 11148)
> /usr/local/pgsql/bin/postgres
>
> and proceeded with:
>
> $ pg_ctl stop -m fast
> waiting for server to shut
> down............................................................... failed
> pg_ctl: server does not shut down
did you try pg_ctl -m immediate stop ???
> > did you try pg_ctl -m immediate stop ??? > > I just did, and it worked. > > $ pg_ctl stop -m immediate > waiting for server to shut down.... done > server stopped I'd be careful about shutting down using "immediate" mode. It forces the database into recovery mode. Your problem could be that one or two connections are in a weird state and even "fast" stopping can't kill them. Next time you have to restart the server, you should check on the status of connections and see if any are in a weird state. I had to deal with this recently where the status was "notify interrupt" and I couldn't even stop fast. I had to change some application code, but not much. Just run "ps ax | grep post" (or whatever options you like to give ps to show all processes) to filter out postgres processes. A connection entry will look like like. [pid] ? Ss 0:00 postgres: [user] [database] [client] [status] Peter
Hello Peter,
Thanks for your post. I appreciate your concern.
> I had to deal with this recently where the
> status was "notify interrupt"
Is this a response from "pg_ctl status" command?
At the moment, I see nothing alarming via ps:
$ ps ax | grep post
11807 ? Ss 0:00 sshd: postgres [priv]
11809 ? S 0:00 sshd: postgres@pts/5
12241 ? S 0:00 /pgsql/pgsql/bin/postgres
12243 ? Ss 2:33 postgres: writer process
12244 ? Ss 0:00 postgres: wal writer process
12245 ? Ss 0:00 postgres: autovacuum launcher process
12246 ? Ss 0:50 postgres: stats collector process
13814 ? SNl 426:30 java -Xms128m -Xmx4600m -jar gadb.jar -loadonly jdbc:db2://172.16.1.109:50001/sci2p mitch winkwink jdbc:postgresql://vixen/canon?prepareThreshold=1 gadb northpole 10
13828 ? Ss 234:43 postgres: gadb canon 127.0.0.1(44824) idle in transaction
13830 ? Ss 235:51 postgres: gadb canon 127.0.0.1(44826) idle in transaction
13832 ? Ss 0:16 postgres: gadb canon 127.0.0.1(44828) idle in transaction
13833 ? Ss 235:03 postgres: gadb canon 127.0.0.1(44829) idle in transaction
13835 ? Ss 234:54 postgres: gadb canon 127.0.0.1(44831) idle
13837 ? Ss 235:48 postgres: gadb canon 127.0.0.1(44833) idle
13839 ? Ss 235:12 postgres: gadb canon 127.0.0.1(44835) idle
13841 ? Rs 234:32 postgres: gadb canon 127.0.0.1(44837) SELECT
13843 ? Ss 234:03 postgres: gadb canon 127.0.0.1(44839) idle in transaction
13845 ? Rs 234:20 postgres: gadb canon 127.0.0.1(44841) SELECT
13847 ? Ss 235:29 postgres: gadb canon 127.0.0.1(44843) idle in transaction
14526 ? Ss 63:15 postgres: ysu canon 127.0.0.1(44853) idle
16563 ? Ss 2:06 postgres: autovacuum worker process canon
21201 pts/8 S+ 0:00 grep post
Regards,
Tena Sakai
tsakai@gallo.ucsf.edu
-----Original Message-----
From: Peter Koczan [mailto:pjkoczan@gmail.com]
Sent: Tue 3/25/2008 10:59 AM
To: Tena Sakai
Cc: Scott Marlowe; pgsql-admin@postgresql.org
Subject: Re: [ADMIN] it refuses to go down...
> > did you try pg_ctl -m immediate stop ???
>
> I just did, and it worked.
>
> $ pg_ctl stop -m immediate
> waiting for server to shut down.... done
> server stopped
I'd be careful about shutting down using "immediate" mode. It forces
the database into recovery mode.
Your problem could be that one or two connections are in a weird state
and even "fast" stopping can't kill them. Next time you have to
restart the server, you should check on the status of connections and
see if any are in a weird state. I had to deal with this recently
where the status was "notify interrupt" and I couldn't even stop fast.
I had to change some application code, but not much.
Just run "ps ax | grep post" (or whatever options you like to give ps
to show all processes) to filter out postgres processes. A connection
entry will look like like.
[pid] ? Ss 0:00 postgres: [user] [database] [client] [status]
Peter
> > I had to deal with this recently where the > > status was "notify interrupt" > > Is this a response from "pg_ctl status" command? I'm referring to the last field of "ps ax". For instance, this line... > 13841 ? Rs 234:32 postgres: gadb canon 127.0.0.1(44837) SELECT would have "notify interrupt" instead of "SELECT". That's the status I'm referring to. Peter
Many thanks, Peter.
Regards,
Tena Sakai
tsakai@gallo.ucsf.edu
-----Original Message-----
From: Peter Koczan [mailto:pjkoczan@gmail.com]
Sent: Thu 3/27/2008 2:59 PM
To: Tena Sakai
Cc: Scott Marlowe; pgsql-admin@postgresql.org
Subject: Re: [ADMIN] it refuses to go down...
> > I had to deal with this recently where the
> > status was "notify interrupt"
>
> Is this a response from "pg_ctl status" command?
I'm referring to the last field of "ps ax". For instance, this line...
> 13841 ? Rs 234:32 postgres: gadb canon 127.0.0.1(44837) SELECT
would have "notify interrupt" instead of "SELECT". That's the status
I'm referring to.
Peter