12.18.4 Shepherd服务

The (gnu services shepherd) module provides a way to define services managed by the GNU Shepherd, which is the initialization system—the first process that is started when the system boots, also known as PID 1 (see Introduction in The GNU Shepherd Manual).

Services in the Shepherd can depend on each other. For instance, the SSH daemon may need to be started after the syslog daemon has been started, which in turn can only happen once all the file systems have been mounted. The simple operating system defined earlier (see 使用配置系统) results in a service graph like this:

You can actually generate such a graph for any operating system definition using the guix system shepherd-graph command (see guix system shepherd-graph).

The %shepherd-root-service is a service object representing PID 1, of type shepherd-root-service-type; it can be extended by passing it lists of <shepherd-service> objects.

Data Type: shepherd-service

The data type representing a service managed by the Shepherd.

provision

This is a list of symbols denoting what the service provides.

These are the names that may be passed to herd start, herd status, and similar commands (see Invoking herd in The GNU Shepherd Manual). See the provides slot in The GNU Shepherd Manual, for details.

requirement (default: '())

List of symbols denoting the Shepherd services this one depends on.

one-shot? (default: #f)

Whether this service is one-shot. One-shot services stop immediately after their start action has completed. See Slots of services in The GNU Shepherd Manual, for more info.

respawn? (default: #t)

Whether to restart the service when it stops, for instance when the underlying process dies.

start

stop (default: #~(const #f))

The start and stop fields refer to the Shepherd’s facilities to start and stop processes (see Service De- and Constructors in The GNU Shepherd Manual). They are given as G-expressions that get expanded in the Shepherd configuration file (see G-表达式).

actions (default: '()) ¶

This is a list of shepherd-action objects (see below) defining actions supported by the service, in addition to the standard start and stop actions. Actions listed here become available as herd sub-commands:

herd action service [arguments…]

auto-start? (default: #t)

Whether this service should be started automatically by the Shepherd. If it is #f the service has to be started manually with herd start.

documentation

A documentation string, as shown when running:

herd doc service-name

where service-name is one of the symbols in provision (see Invoking herd in The GNU Shepherd Manual).

modules (default: %default-modules)

This is the list of modules that must be in scope when start and stop are evaluated.

The example below defines a Shepherd service that spawns syslogd, the system logger from the GNU Networking Utilities (see syslogd in GNU Inetutils):

(let ((config (plain-file "syslogd.conf" "…")))
  (shepherd-service
    (documentation "Run the syslog daemon (syslogd).")
    (provision '(syslogd))
    (requirement '(user-processes))
    (start #~(make-forkexec-constructor
               (list #$(file-append inetutils "/libexec/syslogd")
                     "--rcfile" #$config)
               #:pid-file "/var/run/syslog.pid"))
    (stop #~(make-kill-destructor))))

Key elements in this example are the start and stop fields: they are staged code snippets that use the make-forkexec-constructor procedure provided by the Shepherd and its dual, make-kill-destructor (see Service De- and Constructors in The GNU Shepherd Manual). The start field will have shepherd spawn syslogd with the given option; note that we pass config after --rcfile, which is a configuration file declared above (contents of this file are omitted). Likewise, the stop field tells how this service is to be stopped; in this case, it is stopped by making the kill system call on its PID. Code staging is achieved using G-expressions: #~ stages code, while #$ “escapes” back to host code (see G-表达式).

Data Type: shepherd-action

This is the data type that defines additional actions implemented by a Shepherd service (see above).

名字

Symbol naming the action.

documentation

This is a documentation string for the action. It can be viewed by running:

herd doc service action action

procedure

This should be a gexp that evaluates to a procedure of at least one argument, which is the “running value” of the service (see Slots of services in The GNU Shepherd Manual).

The following example defines an action called say-hello that kindly greets the user:

(shepherd-action
  (name 'say-hello)
  (documentation "Say hi!")
  (procedure #~(lambda (running . args)
                 (format #t "Hello, friend! arguments: ~s\n"
                         args)
                 #t)))

Assuming this action is added to the example service, then you can do:

# herd say-hello example
Hello, friend! arguments: ()
# herd say-hello example a b c
Hello, friend! arguments: ("a" "b" "c")

This, as you can see, is a fairly sophisticated way to say hello. See Service Convenience in The GNU Shepherd Manual, for more info on actions.

Scheme Procedure: shepherd-configuration-action

Return a configuration action to display file, which should be the name of the service’s configuration file.

It can be useful to equip services with that action. For example, the service for the Tor anonymous router (see tor-service-type) is defined roughly like this:

(let ((torrc (plain-file "torrc" …)))
  (shepherd-service
    (provision '(tor))
    (requirement '(user-processes loopback syslogd))

    (start #~(make-forkexec-constructor
              (list #$(file-append tor "/bin/tor") "-f" #$torrc)
              #:user "tor" #:group "tor"))
    (stop #~(make-kill-destructor))
    (actions (list (shepherd-configuration-action torrc)))
    (documentation "Run the Tor anonymous network overlay.")))

Thanks to this action, administrators can inspect the configuration file passed to tor with this shell command:

cat $(herd configuration tor)

This can come in as a handy debugging tool!

Scheme Variable: shepherd-root-service-type

The service type for the Shepherd “root service”—i.e., PID 1.

This is the service type that extensions target when they want to create shepherd services (see 服务类型和服务, for an example). Each extension must pass a list of <shepherd-service>. Its value must be a shepherd-configuration, as described below.

Data Type: shepherd-configuration

This data type represents the Shepherd’s configuration.

shepherd (default: shepherd)

The Shepherd package to use.

services (default: '())

A list of <shepherd-service> to start. You should probably use the service extension mechanism instead (see Shepherd服务).

The following example specifies the Shepherd package for the operating system:

(operating-system
  ;; ...
  (services (append (list openssh-service-type))
            ;; ...
            %desktop-services)
  ;; ...
  ;; Use own Shepherd package.
  (essential-services
   (modify-services (operating-system-default-essential-services
                     this-operating-system)
     (shepherd-root-service-type config => (shepherd-configuration
                                            (inherit config)
                                            (shepherd my-shepherd))))))

Scheme Variable: %shepherd-root-service

This service represents PID 1.