repmgrd starting continuous standby node monitoring

TODO

@@ -0,0 +1 @@
+fix stop_standby

manifests/config.pp

@@ -28,7 +28,7 @@
     $pg_contribdir = '/usr/share/postgresql/9.1/contrib'
 
     $repmgr_config_file = '/etc/repmgr/repmgr.conf'
-    #$repmgr_opts = join(['--force', '--verbose'], ' ')
+    $repmgrd_logdir = '/var/log/repmgr'
     $repmgr_opts = join(['--force', '--verbose'], ' ')
 
     Account['postgres'] -> File['repmgr_config_file'] -> File['pg_ssh_config']
@@ -51,6 +51,15 @@
         path      => $repmgr_config_file,
     }
 
+    # Assert that repmgrd log directory exists
+    file {'repmgrd_log_dir':
+        ensure => directory,
+        path   => $repmgrd_logdir,
+        owner  => postgres,
+        group  => postgres,
+        mode   => 644,
+    }
+
     # Assert postgres user exists
     account {'postgres':
         ensure => present,
@@ -71,25 +80,25 @@
         content => 'StrictHostKeyChecking no',
     }
 
-
     if $node_role == 'master' {
 
-        # Create the user and database to manage replication
         Exec['create_repmgr_db_user'] -> Exec['create_repmgr_db'] -> Exec['master_register']
 
+        # Create the user and database to manage replication
         exec {'create_repmgr_db_user':
             path     => ['/bin/','/usr/bin'],
             command  => 'createuser -s repmgr',
             user     => 'postgres',
             onlyif   => '[ `psql template1 -c "\du" | grep -c repmgr` -eq "0" ]',
         }
+
         exec{'create_repmgr_db':
             path     => ['/bin', '/usr/bin'],
             command  => "createdb -O repmgr repmgr && psql -f $pg_contribdir/repmgr_funcs.sql repmgr",
             user     => 'postgres',
             onlyif   => '[ `psql -l | grep -c repmgr` -eq "0" ]',
-        }
-
+        } 
+        # Register master node if it's not and only if repmgr database has been created
         exec {'master_register':
             path    => ['/bin', '/usr/bin'],
             command => "repmgr -f $repmgr_config_file master register",
@@ -102,22 +111,32 @@
     }
     elsif $node_role in ['slave', 'witness'] {
 
-        Exec['stop_standby'] -> Exec['clone_master']
+        Exec['stop_standby'] -> Exec['clone_master'] ~> Service['postgresql'] -> Exec['standby_register'] -> Exec['start_repmgrd']
 
         # Stop standby node if running
         exec {'stop_standby':
             path => ['/usr/bin', '/usr/lib/postgresql/9.1/bin'],
-            command => "sudo -u postgres $pg_ctl -D $pg_data stop -l $pg_logfile",
-            onlyif  => "[ killall -0 postgres | grep -c postgres` -eq 0 ]",
+            command => "sudo -u postgres $pg_ctl -D $pg_data stop -m fast -l $pg_logfile",
+            onlyif  => "[ `killall -0 postgres 2>&1 | grep -c postgres` -eq 0 ]",
         }
         # Clone the master and start standby
         exec {'clone_master':
             path    => ['/usr/bin'],
             command => "sudo -u postgres repmgr -f $repmgr_config_file -D $pg_data -d repmgr -U repmgr -R postgres standby clone $master_node $repmgr_opts",
-            notify  => Service['postgresql'],
+            #notify  => Service['postgresql'],
+        }
+        # Register the standby server if it's not
+        exec {'standby_register':
+            path    => ['/bin', '/usr/bin'],
+            command => "sudo -u postgres repmgr -f $repmgr_config_file standby register",
+            onlyif  => "[ `repmgr -f $repmgr_config_file cluster show | grep -c 'standby | host=$node_name'` -eq 0 ]"
         }
-        # Register the standby server
         # Start repmgrd daemon
+        # NOTE running repmgrd via an init script will be better
+        exec {'start_repmgrd':
+            path    => ['/bin', '/usr/bin'],
+            command => "sudo -u postgres repmgrd -f $repmgr_config_file -d --monitoring-history > ${repmgrd_logdir}/repmgrd.log 2>&1",
+            onlyif   => "[ `killall -0 repmgrd 2>&1 | grep -c repmgrd` -ne 0 ]",
+        }
     }
-
 }

manifests/postgresql.pp

@@ -57,7 +57,7 @@
         start   => "sudo -u postgres $pg_ctl -D $pg_data start -o '-c config_file=$pg_configdir/postgresql.conf' -l $pg_logfile && sleep 10",
         # The real Debian way is to use pg_ctlcluster like so :
         # sudo -u postgres pg_ctlcluster 9.3 main start 
-        stop    => "sudo -u postgres $pg_ctl -D $pg_data stop -l $pg_logfile",
+        stop    => "sudo -u postgres $pg_ctl -D $pg_data stop -m fast -l $pg_logfile",
         restart => "sudo -u postgres $pg_ctl -D $pg_data restart -l $pg_logfile && sleep 10",
     }