start

This command starts one or more services.

Interface

start [ -h ] [ -P ] service...

This command expects to find an already running scandir. If the state of the service is already up, the start command does nothing but handles dependency except when passing the -P option. A classic service already up receives the order again and stays up: its state does not change, so no up event is sent and no reactor fires on it. A oneshot, a module or an event source already up is skipped, said at verbosity 2.

Multiple services can be started by separating their names with a space.

If the service has never been parsed, the start command will parse it and associate the service with a specific tree if the 66 -t option was used, or if the frontend file defines the InTree field, or if a tree was marked as current—refer to the tree current subcommand. Otherwise, the service will be associated with the global tree. This process is applied to all dependencies of the service. A frontend the parse refuses starts nothing, see parse.

If the service was already parsed, the 66 -t command option has no effect.

In case of module service type, all services declared within the module will start, in their order, and the module is marked up after them. The -P has no effect on the services within the module and only affects the module's dependencies.

A logger is started with its service. If a service conflicts with a running one through its Conflict key, nothing starts.

If the service is a reactor whose action is Do = start or Do = restart, it is armed: its rule is registered at 66-eventd and its Execute does not run until its event comes. A reactor with another action starts as any service. The sources named in its From come up first, except under -P.

A service fails to start when its script fails or when it is not ready in time, and what waits for it is not started. The command then exits with 111 and says, at verbosity 2, how the service ended:

unable to start service: <name> -- exited with code <n>
unable to start service: <name> -- killed by signal <n>
unable to start service: <name> -- not ready within its start timeout
unable to start service: <name> -- a service it waits for failed

A classic service that declares Notify has its [Start] Timeout to become ready, counted from the start of its process. When the time runs out, its supervisor brings it down, it stays down and supervised, and it is not started again until the next start.

This command handles interdependencies.

Options

• -h, --help: prints this help.

• -P, --no-propagate: Do not handle service dependencies. In such cases, the start command will not attempt to start the services that the service depends on, regardless of their current state. The named services only are started, not their logger. A named service never parsed is parsed alone: its dependencies never parsed are not parsed, and stay absent until a command without -P parses them.

Usage examples

Start a service handling its dependencies.

66 start foo

Start the service without handling any dependencies

66 start -P foo

Start a service in a specific tree if the service was never parsed

66 -t treeA start foo

Modules services

Services within a module can also be managed independently. If you need to start a particular service inside the module, specify the name of the module service followed by a colon : and the name of the service within the module

66 start foo@foobar:foobaz

where foo@foobar is the name of the module service and foobaz the name of the service inside the module service.

You also can use the -P option to avoid handling the dependencies of the service inside the module service

66 start -P foo@foobar:foobaz