Kill stale pidfiles before starting the service

If you kill postgres badly, it leaves a `postmaster.pid` file lying
around which prevents the service from starting later. This gets in the
way of development, causes confusion, and *burns battery* as launchd
tries again and again to start postgres.

There are potential consequences to this in the case of a corrupt
database, but they're so unlikely that I much prefer to err on the side
of convenience for a developer machine.

Author: dgoodlad

manifests/service.pp

@@ -16,6 +16,7 @@
   }
 
   $nc = "nc -z ${host} ${port}"
+  $pidfile = "${datadir}/postmaster.pid"
 
   if $::operatingsystem == 'Darwin' {
     service { 'com.boxen.postgresql':
@@ -30,6 +31,13 @@
     creates => "${datadir}/PG_VERSION",
   }
 
+  ->
+  exec { 'kill-stale-postgres-pidfile':
+    # Remove the pidfile, unless there's a postgres process running
+    command => "rm -f ${pidfile}",
+    unless  => "ps -p `head -1 ${pidfile}` | grep postgres",
+  }
+
   ->
   service { $service:
     ensure => $svc_ensure,