postgresql-9.5 on CentOS 7 won't start on boot (PGDATA on mount point)

Question

I am trying to get postgresql-9.5 to start on boot on CentOS 7.

I've followed the installation and startup steps here https://wiki.postgresql.org/wiki/YUM_Installation And have tried some additional troubleshooting from this question here https://stackoverflow.com/questions/35492893/unable-to-start-postgresql-9-5-on-centos-7

It might be relevant that I am using a custom PGDATA directory that is a mount point. (/mnt/data)

The database was initialized pointing to the /mnt/data directory.

su postgres -c "/usr/pgsql-9.5/bin/pg_ctl initdb -D /mnt/data"

I made the necessary changes to the /usr/lib/systemd/system/postgresql-9.5.service file to point to the /mnt/data directory.

The service was enabled and started successfully.

systemctl enable postgresql-9.5

systemctl start postgresql-9.5

After rebooting, systemctl status postgresql-9.5 shows the following:

● postgresql-9.5.service - PostgreSQL 9.5 database server
   Loaded: loaded (/usr/lib/systemd/system/postgresql-9.5.service; enabled; vendor preset: disabled)
   Active: failed (Result: exit-code) since Wed 2016-04-06 00:53:02 EDT; 4min 36s ago
  Process: 9911 ExecStartPre=/usr/pgsql-9.5/bin/postgresql95-check-db-dir ${PGDATA} (code=exited, status=1/FAILURE)

Apr 06 00:53:02 pgdev-fl1 systemd[1]: Starting PostgreSQL 9.5 database server...
Apr 06 00:53:02 pgdev-fl1 systemd[1]: postgresql-9.5.service: control process exited, code=exited status=1
Apr 06 00:53:02 pgdev-fl1 systemd[1]: Failed to start PostgreSQL 9.5 database server.
Apr 06 00:53:02 pgdev-fl1 systemd[1]: Unit postgresql-9.5.service entered failed state.
Apr 06 00:53:02 pgdev-fl1 systemd[1]: postgresql-9.5.service failed.

However, I can immediately run systemctl start postgresql-9.5 and it loads just fine, using the /mnt/data directory.

Thinking it might be an issue of timing and it trying to start the service before the mount point exists, I included Requires=mnt-data.mount in the [Unit] section of /usr/lib/systemd/system/postgresql-9.5.service, and re-enabled the service after a systemctl daemon-reload.

Yet still it errors out with the same message upon boot.

Running /usr/pgsql-9.5/bin/postgresql95-check-db-dir ${PGDATA} generates no output for me so I am at a loss as to where to continue troubleshooting this issue.

Any guidance is appreciated!

Update - 4/6/16

After checking journalctl it's apparent that it is trying to load the postgresql-9.5 service before the iSCSI volume is mounted, even though the Requires directive was included.

Answer 1

After rereading https://www.digitalocean.com/community/tutorials/understanding-systemd-units-and-unit-files it's become apparent to me that Requires doesn't behave how I thought it would.

Changing the directive to After solved the issue and made the postgresql-9.5 service wait until the mount point was available.

Requires without the After will try and load them simultaneously.

You can use Requires with After to imply the dependency relationship.