Postgresql Cost-Based Vacuum Delay parameters

The fact that you have no lines in pg_stat_replication on the master indicates that your streaming replication is not working - and it has instead fallen back on the file based replication, which has more delay.

Check your logs on both master and slave, they should tell you why it's not starting up in streaming mode.

Also, your recovery.conf file is incorrect - you should not be using pg_standby together with standby_mode. That's a leftover from pre-9.0 days.

You are not doing anything wrong per se, and the documentation is not wrong. But the initscript from CentOS doesn't support you doing things this way, so you have to change the initscripts to not check for these files, since PostgreSQL will check for it itself and just try to start the database even if these files are not there.

So to test it out, comment out the if, then and the whole else block and just let it run the ... part of the script. PostgreSQL should read the settings correctly and start itself without a problem. And you might of course inform the people responsible for the CentOS package that they should do it differently. If it doesn't run, you should get a reasonable error message from PostgreSQL itself.