Skip to content

chef-cookbooks/runit

Repository files navigation

runit Cookbook

Build Status Cookbook Version

Installs runit and provides the runit_service service resource for managing processes (services) under runit.

This cookbook does not use runit to replace system init, nor are there plans to do so.

For more information about runit:

NOTE: The 5.0 release of this cookbook requires the ChefSpec which shipped in the later versions of ChefDK 3. If you use this cookbook along with ChefSpec in your environment then you will need to upgrade to the latest version of ChefDK / Workstation to prevent spec failures.

Requirements

Platforms

  • Debian/Ubuntu
  • RHEL and derivatives

Chef

  • Chef 14.0+

Cookbooks

  • packagecloud (for RHEL)
  • yum-epel (for RHEL)

Attributes

  • node['runit']['prefer_local_yum'] - If true, assumes that a runit package is available on an already configured local yum repository. By default, the recipe installs the runit package from a Package Cloud repository (see below). This is set to the value of node['runit']['use_package_from_yum'] for backwards compatibility, but otherwise defaults to false.

Recipes

default

The default recipe installs runit and starts runsvdir to supervise the services in runit's service directory (e.g., /etc/service).

On RHEL-family systems, it will install the runit RPM using Ian Meyer's Package Cloud repository for runit. This replaces the previous functionality where the RPM was build using his runit RPM SPEC. However, if the attribute node['runit']['prefer_local_yum'] is set to true, the packagecloud repository creation will be skipped and it is assumed that a runit package is available on an otherwise configured (outside this cookbook) local repository.

On Debian family systems, the runit packages are maintained by the runit author, Gerrit Pape, and the recipe will use that for installation.

Resource

This cookbook has a resource, runit_service, for managing services under runit.

Actions

  • enable - enables the service, creating the required run scripts and symlinks. This is the default action.
  • start - starts the service with sv start
  • stop - stops the service with sv stop
  • disable - stops the service with sv down and removes the service symlink
  • create - create the service directory, but don't enable the service with symlink
  • restart - restarts the service with sv restart
  • reload - reloads the service with sv force-reload
  • reload_log - reloads the service's log service
  • once - starts the service with sv once.
  • hup - sends the HUP signal to the service with sv hup
  • cont - sends the CONT signal to the service
  • term - sends the TERM signal to the service
  • kill - sends the KILL signal to the service
  • up - starts the service with sv up
  • down - downs the service with sv down
  • usr1 - sends the USR1 signal to the service with sv 1
  • usr2 - sends the USR2 signal to the service with sv 2

Service management actions are taken with runit's "sv" program.

Read the sv(8) man page for more information on the sv program.

Properties

The first three properties, sv_dir, service_dir, and sv_bin will attempt to use the legacy node attributes, and fall back to hardcoded default values that match the settings used on Debian platform systems.

Many of these properties are only used in the :enable action.

  • sv_dir - The base "service directory" for the services managed by the resource. By default, this will attempt to use the node['runit']['sv_dir'] attribute, and falls back to /etc/sv.
  • service_dir - The directory where services are symlinked to be supervised by runsvdir. By default, this will attempt to use the node['runit']['service_dir'] attribute, and falls back to /etc/service.
  • lsb_init_dir - The directory where an LSB-compliant init script interface will be created. By default, this will attempt to use the node['runit']['lsb_init_dir'] attribute, and falls back to /etc/init.d.
  • sv_bin - The path to the sv program binary. This will attempt to use the node['runit']['sv_bin'] attribute, and falls back to /usr/bin/sv.
  • service_name - Name attribute. The name of the service. This will be used in the directory of the managed service in the sv_dir and service_dir.
  • sv_timeout - Override the default sv timeout of 7 seconds.
  • sv_verbose - Whether to enable sv verbose mode. Default is false.
  • sv_templates - If true, the :enable action will create the service directory with the appropriate templates. Default is true. Set this to false if the service has a package that provides its own service directory. See Usage examples.
  • options - DEPRECATED - Options passed as variables to templates, for compatibility with legacy runit service definition. Default is an empty hash.
  • env - A hash of environment variables with their values as content used in the service's env directory. Default is an empty hash. When this hash is non-empty, the contents of the runit service's env directory will be managed by Chef in order to conform to the declared state.
  • log - Whether to start the service's logger with svlogd, requires a template sv-service_name-log-run.erb to configure the log's run script. Default is true.
  • default_logger - Whether a default log/run script should be set up. If true, the default content of the run script will use svlogd to write logs to /var/log/service_name. Default is false.
  • log_dir - The directory where the svlogd log service will run. Used when default_logger is true. Default is /var/log/service_name
  • log_flags - The flags to pass to the svlogd command. Used when default_logger is true. Default is -tt
  • log_size - The maximum size a log file can grow to before it is automatically rotated. See svlogd(8) for the default value.
  • log_num - The maximum number of log files that will be retained after rotation. See svlogd(8) for the default value.
  • log_min - The minimum number of log files that will be retained after rotation (if svlogd cannot create a new file and the minimum has not been reached, it will block). Default is no minimum.
  • log_timeout - The maximum age a log file can get to before it is automatically rotated, whether it has reached log_size or not. Default is no timeout.
  • log_processor - A string containing a path to a program that rotated log files will be fed through. See the PROCESSOR section of svlogd(8) for details. Default is no processor.
  • log_socket - An string containing an IP:port pair identifying a UDP socket that log lines will be copied to. Default is none.
  • log_prefix - A string that will be prepended to each line as it is logged. Default is no prefix.
  • log_config_append - A string containing optional additional lines to add to the log service configuration. See svlogd(8) for more details.
  • cookbook - A cookbook where templates are located instead of where the resource is used. Applies for all the templates in the enable action.
  • check - whether the service has a check script, requires a template sv-service_name-check.erb
  • finish - whether the service has a finish script, requires a template sv-service_name-finish.erb
  • control - An array of signals to customize control of the service, see runsv man page on how to use this. This requires that each template be created with the name sv-service_name-signal.erb.
  • supervisor_owner - the user that should be allowed to control this service, see runsv faq
  • supervisor_group - the group that should be allowed to control this service, see runsv faq
  • owner - user that should own the templates created to enable the service
  • group - group that should own the templates created to enable the service
  • run_template_name - alternate filename of the run run script to use replacing service_name.
  • log_template_name - alternate filename of the log run script to use replacing service_name.
  • check_script_template_name - alternate filename of the check script to use, replacing service_name.
  • finish_script_template_name - alternate filename of the finish script to use, replacing service_name.
  • control_template_names - a hash of control signals (see control above) and their alternate template name(s) replacing service_name.
  • status_command - The command used to check the status of the service to see if it is enabled/running (if it's running, it's enabled). This hardcodes the location of the sv program to /usr/bin/sv due to the aforementioned cookbook load order.
  • restart_on_update - Whether the service should be restarted when the run script is updated. Defaults to true. Set to false if the service shouldn't be restarted when the run script is updated.
  • start_down - Set the default state of the runit service to 'down' by creating <sv_dir>/down file. Defaults to false. Services using start_down will not be notified to restart when their run script is updated.
  • delete_downfile - Delete previously created <sv_dir>/down file

Unlike previous versions of the cookbook using the runit_service definition, the runit_service resource can be notified. See Usage examples below.

Usage

To get runit installed on supported platforms, use recipe[runit]. Once it is installed, use the runit_service resource to set up services to be managed by runit.

In order to use the runit_service resource in your cookbook(s), each service managed will also need to have sv-service_name-run.erb and sv-service_name-log-run.erb templates created. If the log property is false, the log run script isn't created. If the log property is true, and default_logger is also true, the log run script will be created with the default content:

#!/bin/sh
exec svlogd -tt /var/log/service_name

Examples

These are example use cases of the runit_service resource described above. There are others in the runit_test cookbook that is included in the git repository.

Default Example

This example uses all the defaults in the :enable action to set up the service.

We'll set up chef-client to run as a service under runit, such as is done in the chef-client cookbook. This example will be more simple than in that cookbook. First, create the required run template, chef-client/templates/default/sv-chef-client-run.erb.

#!/bin/sh
exec 2>&1
exec /usr/bin/env chef-client -i 1800 -s 30

Then create the required log/run template, chef-client/templates/default/sv-chef-client-log-run.erb.

#!/bin/sh
exec svlogd -tt ./main

Note This will cause output of the running process to go to /etc/sv/chef-client/log/main/current. Some people may not like this, see the following example. This is preserved for compatibility reasons.

Finally, set up the service in the recipe with:

runit_service "chef-client"

Default Logger Example

To use a default logger with svlogd which will log to /var/log/chef-client/current, instead, use the default_logger option.

runit_service "chef-client" do
  default_logger true
end

No Log Service

If there isn't an appendant log service, set log to false, and the log/run script won't be created.

runit_service "no-svlog" do
  log false
end

Check Script

To create a service that has a check script in its service directory, set the check property to true, and create a sv-checker-check.erb template.

runit_service "checker" do
  check true
end

This will create /etc/sv/checker/check.

Finish Script

To create a service that has a finish script in its service directory, set the finish property to true, and create a sv-finisher-finish.erb template.

runit_service "finisher" do
  finish true
end

This will create /etc/sv/finisher/finish.

Alternate service directory

If the service directory for the managed service isn't the sv_dir (/etc/sv), then specify it:

runit_service "custom_service" do
  sv_dir "/etc/custom_service/runit"
end

No Service Directory

If the service to manage has a package that provides its service directory, such as git-daemon on Debian systems, set sv_templates to false.

package "git-daemon-run"

runit_service "git-daemon" do
  sv_templates false
end

This will create the service symlink in /etc/service, but it will not manage any templates in the service directory.

User Controlled Services

To set up services controlled by a non-privileged user, we follow the recommended configuration in the runit documentation (Is it possible to allow a user other than root to control a service?).

Suppose the user's name is floyd, and floyd wants to run floyds-app. Assuming that the floyd user and group are already managed with Chef, create a runsvdir-floyd runit_service.

runit_service "runsvdir-floyd"

Create the sv-runsvdir-floyd-log-run.erb template, or add log false. Also create the sv-runsvdir-floyd-run.erb with the following content:

#!/bin/sh
exec 2>&1
exec chpst -ufloyd runsvdir /home/floyd/service

Next, create the runit_service resource for floyd's app:

runit_service "floyds-app" do
  sv_dir "/home/floyd/sv"
  service_dir "/home/floyd/service"
  owner "floyd"
  group "floyd"
end

And now floyd can manage the service with sv:

$ id
uid=1000(floyd) gid=1001(floyd) groups=1001(floyd)
$ sv stop /home/floyd/service/floyds-app/
ok: down: /home/floyd/service/floyds-app/: 0s, normally up
$ sv start /home/floyd/service/floyds-app/
ok: run: /home/floyd/service/floyds-app/: (pid 5287) 0s
$ sv status /home/floyd/service/floyds-app/
run: /home/floyd/service/floyds-app/: (pid 5287) 13s; run: log: (pid 4691) 726s

Options

Next, let's set up memcached under runit with some additional options using the options property. First, the memcached/templates/default/sv-memcached-run.erb template:

#!/bin/sh
exec 2>&1
exec chpst -u <%= @options[:user] %> /usr/bin/memcached -v -m <%= @options[:memory] %> -p <%= @options[:port] %>

Note that the script uses chpst (which comes with runit) to set the user option, then starts memcached on the specified memory and port (see below).

The log/run template, memcached/templates/default/sv-memcached-log-run.erb:

#!/bin/sh
exec svlogd -tt ./main

Finally, the runit_service in our recipe:

runit_service "memcached" do
  options({
    :memory => node[:memcached][:memory],
    :port => node[:memcached][:port],
    :user => node[:memcached][:user]
  }.merge(params))
end

This is where the user, port and memory options used in the run template are used.

Notifying Runit Services

In previous versions of this cookbook where the definition was used, it created a service resource that could be notified. With the runit_service resource, recipes need to use the full resource name.

For example:

runit_service "my-service"

template "/etc/my-service.conf" do
  notifies :restart, "runit_service[my-service]"
end

Because the resource implements actions for various commands that sv can send to the service, any of those actions could be used for notification. For example, chef-client supports triggering a Chef run with a USR1 signal.

template "/tmp/chef-notifier" do
  notifies :usr1, "runit_service[chef-client]"
end

For older implementations of services that used runit_service as a definition, but may support alternate service styles, use a conditional, such as based on an attribute:

service_to_notify = case node['nginx']['init_style']
                    when "runit"
                      "runit_service[nginx]"
                    else
                      "service[nginx]"
                    end

template "/etc/nginx/nginx.conf" do
  notifies :restart, service_to_notify
end

More Examples

For more examples, see the runit_test cookbook's service recipe in the git repository.

Development

You may use test kitchen with either the vagrant or docker drivers to run integration tests.

Note: When using the docker driver please ensure that the container you are using has a working init system, as runit expects to be started by init. In some cases, systemd may need to be run in privileged mode.

For instance, for ubuntu with upstart:

    driver_config:
      image: ubuntu-upstart:14.04
      run_command: /sbin/init

For redhat derivatives:

    driver_config:
      image: dockerhub/image-with-systemd
      run_command: /usr/sbin/init
      privileged: true

License & Authors

Copyright:: 2008-2019, Chef Software, Inc

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.