snapshot of https://github.com/zulip/zulip.git @ 48c69911c7dd4e03b20552eb3bd77ad8151c1357 (2026-09-15)

puppet/zulip/manifests/supervisor.pp

@@ -0,0 +1,101 @@
+class zulip::supervisor {
+  $supervisor_service = $zulip::common::supervisor_service
+
+  package { 'supervisor': ensure => installed }
+
+  $system_conf_dir = $zulip::common::supervisor_system_conf_dir
+  file { $system_conf_dir:
+    ensure  => directory,
+    require => Package['supervisor'],
+    owner   => 'root',
+    group   => 'root',
+  }
+
+  $conf_dir = $zulip::common::supervisor_conf_dir
+  file { $conf_dir:
+    ensure  => directory,
+    require => Package['supervisor'],
+    owner   => 'root',
+    group   => 'root',
+    purge   => true,
+    recurse => true,
+    notify  => Service[$supervisor_service],
+  }
+
+  # In the docker environment, we don't want/need supervisor to be
+  # started/stopped /bin/true is used as a decoy command, to maintain
+  # compatibility with other code using the supervisor service.
+  $puppet_classes = zulipconf('machine', 'puppet_classes', undef)
+  if 'docker' in $puppet_classes {
+    service { $supervisor_service:
+      ensure     => running,
+      require    => [
+        File['/var/log/zulip'],
+        Package['supervisor'],
+      ],
+      hasstatus  => true,
+      status     => '/bin/true',
+      hasrestart => true,
+      restart    => '/bin/true',
+    }
+    exec { 'supervisor-restart':
+      refreshonly => true,
+      command     => '/bin/true',
+    }
+  } else {
+    service { $supervisor_service:
+      ensure     => running,
+      require    => [
+        File['/var/log/zulip'],
+        Package['supervisor'],
+      ],
+      hasstatus  => true,
+      status     => $zulip::common::supervisor_status,
+      # Restarting the whole supervisorctl on every update to its
+      # configuration files has the unfortunate side-effect of
+      # restarting all of the services it controls; this results in an
+      # unduly large disruption.  The better option is to tell
+      # supervisord to reread its config via supervisorctl and then to
+      # "update".  You need to do both -- after a "reread", supervisor
+      # won't actually take action based on the changed configuration
+      # until you do an "update" (I assume this is so you can check if
+      # your config file parses without doing anything, but it's
+      # really confusing).
+      #
+      # If restarting supervisor itself is necessary, see
+      # Exec['supervisor-restart']
+      #
+      # Also, to handle the case that supervisord wasn't running at
+      # all, we check if it is not running and if so, start it.
+      hasrestart => true,
+      # lint:ignore:140chars
+      restart    => "bash -c 'if pgrep -x supervisord >/dev/null; then supervisorctl reread && supervisorctl update; else ${zulip::common::supervisor_start}; fi'",
+      # lint:endignore
+    }
+    exec { 'supervisor-restart':
+      refreshonly => true,
+      provider    => shell,
+      command     => $zulip::common::supervisor_reload,
+      require     => Service[$supervisor_service],
+    }
+  }
+
+  $file_descriptor_limit = zulipconf('application_server', 'service_file_descriptor_limit', 40000)
+  file { $zulip::common::supervisor_conf_file:
+    ensure  => file,
+    require => Package[supervisor],
+    owner   => 'root',
+    group   => 'root',
+    mode    => '0644',
+    content => template('zulip/supervisor/supervisord.conf.erb'),
+    notify  => Exec['supervisor-restart'],
+  }
+
+  file { '/usr/local/bin/secret-env-wrapper':
+    ensure => file,
+    owner  => 'root',
+    group  => 'root',
+    mode   => '0755',
+    source => 'puppet:///modules/zulip/secret-env-wrapper',
+  }
+}