Re: Problem changing default data_directory in PG 9.6 + CentOS6

On 10/17/2016 01:38 PM, Edilmar LISTAS wrote:
> I have an env running a changed data_directory fine in a devel machine
> PG 9.4 using Fedora23.
> Now, I have a server machine with CentOS where I downloaded the RPMs
> from repo
> https://download.postgresql.org/pub/repos/yum/9.6/redhat/rhel-6-x86_64/pgdg-centos96-9.6-3.noarch.rpm.
> All the configs run fine if I doesn't change the default data_directory.
> But I need to use the path /sistemas/sat4/bdpg.
>
> I did these commands:
>
> mkdir /sistemas/sat4/bdpg
> chown postgres /sistemas/sat4/bdpg
> chmod 700 /sistemas/sat4/bdpg
> su - postgres
> /usr/pgsql-9.6/bin/initdb -D /sistemas/sat4/bdpg
> exit
>
> Then, I changed data_directory to /sistemas/sat4/bdpg and tried to

Changed data_directory where?

> restart PG:
> service postgresql-9.6 restart
> STOP => OK
> START => FAILED
>
> I disabled se_linux.
> The file /var/lib/pgsql/9.6/pgstartup.log just said to see future output
> in pg_log.
> The file data/pg_log/postgresql-Mon.log doesn't say anything about errors.
>
> The strange is that startup arises a FAILED message, but the process
> "/usr/pgsql-9.6/bin/postmaster -D /var/lib/pgsql/9.6/data" is running

So /var/lib/pgsql/9.6/data is where the original install is?

Best guess is that some script is starting the original install and when
you go to start your custom location it fails because the new cluster is
trying to use the port(5432 I am assuming) as the original cluster.

Have you tried giving the new cluster a different port number, say 5442,
and the starting it?


> (and the children logger/checkpointer/etc). But I don't get to connect
> using pgAdmin3. Then, I have to kill manually postmaster (because
> service script doesn't understand postmaster.pid in the new data dir),
> comment data_directory to use default place, start and connect to
> pgAdmin3. Then, start/stop/start/etc run fine lookup for postmaster.pid
> in /var/lib/pgsql/9.6/data.

So you either need to change the start script to point to the new
cluster or create a new one for it.



--
Adrian Klaver
adrian.klaver@aklaver.com