Install & configure Supervisor. This role provides it's own Upstart init script, and can install it for each supervised application. It will also deactivate the distro's one (configurable with supervisord_enable_distro_init).
Pretty static ansible_managed.
This role uses the ansible_managed
variable. If you set it to something dynamic (or left the default), you should pass a static one into this role - otherwise you'll see unnecessary restarts.
None of them are required, but you probably want to set supervisord_programs
and supervisord_instance_name
.
Escape warning: most options need "%" escaped as "%%", as supervisord uses %(var_name)s for variable substitution.
supervisord_instance_name: "default"
This name will be used in default paths, and for init script names (this will be the name of the created system service). If you need multiple supervisor daemons, that's how they are told apart.
supervisord_configuration_file: "/etc/supervisord-{{ supervisord_instance_name }}.conf"
Where to put this instance's configuration.
supervisord_logfile: "/var/log/supervisor/{{ supervisord_instance_name }}.log"
Main log file location.
supervisord_user: no
Run supervisord as this user. This is set at supervisord level, not upstart level.
supervisord_kill_timeout: 60
Time (in seconds) between sending SIGTERM and SIGKILL when stopping supervisord.
supervisord_environment: {}
Environment for supervisord and all children. This can have 2 forms, either a string (that will be put literally into config file) or a dictionary of variable: value.
supervisord_identifier: "{{ supervisord_instance_name }}"
Identifier for use in RPC communication.
supervisord_programs: []
Programs this instance should supervise. List of dictionaries, fields used here are described in Program Definition section.
supervisord_disable_distro_init: yes
Should the distribution-provided init script be turned off? This will also stop the instance managed by that script.
supervisord_extra_config: {}
Additional configuration for supervisord.
It's a dictionary mapping section name to a dictionary mapping options to values, for example: { inet_http_server: { port: 9001 } }
.
If some option is also set by this role, it might show up twice (so be careful),
but sections won't be duplicated.
Each element of supervisord_programs defines a process group. The full list of settings is at: http://supervisord.org/configuration.html#program-x-section-values. Most settings are just written into the file (with some care for booleans), this lists only includes the special ones.
name: No default, required
Name of this process group. This will be available as %(program_name)s in all string values in this group.
command: No default, required
The command to run.
socket: No default, optional
Socket to listen on, on behalf of this program. Setting this will generate an fcgi-program:{{ name }} instead of program:{{ name }} section.
environment: {}
Additional environment variables for this process group. For format see supervisord_environment in Role Variables.
This role exports following handlers:
restart {{ supervisord_instance_name }}
Restarts the whole supervisord process.
reload {{ supervisord_instance_name }}
Reloads supervisord configuration.
When using this role as a dependency, you'll probably want to pass your role's name as supervisord_instance_name and use something like this for handlers:
- name: Restart my web workers
supervisorctl:
name: my_app-web
state: restarted
config: "{{ supervisord_configuration_file }}"
One warning though - if you use supervisorctl with state=started just after reloaading/restarting supervisord
it might throw strange looking errors like ERROR (already started)
.
This is a bug in supervisorctl module, for a workaround see _vagrant/playbook.yml Proposed fix to this bug is here: ansible/ansible#7980
- hosts: my_app_servers
roles:
- my_app_role
- role: ktosiek.supervisord
supervisord_instance_name: "my_app"
supervisord_programs:
- name: my_app
command: /opt/my_app/fcgi
socket: /var/run/my_app/my_app.socket
BSD