initctl(8) System Manager's Manual initctl(8)
NAME
initctl - init daemon control tool
SYNOPSIS
initctl [OPTION]... COMMAND [OPTION]... ARG...
DESCRIPTION
initctl allows a system administrator to communicate and interact with the Upstart init(8) daemon.
If D-Bus has been configured to allow non-privileged users to invoke all Upstart D-Bus methods, this command is also able to manage user
jobs. See init(5) for further details.
When run as initctl, the first non-option argument is the COMMAND. Global options may be specified before or after the command.
You may also create symbolic or hard links to initctl named after commands. When invoked through these links the tool will behave only as
that command, with global and command-specific options intermixed. The default installation supplies such links for the start, stop,
restart, reload and status commands.
OPTIONS
--session
Connect to init(8) daemon using the D-Bus session bus (for testing purposes only).
--system
Communication with the init(8) daemon is normally performed over a private socket connection. This has the advantage of speed and
robustness, when issuing commands to start or stop services or even reboot the system you do not want to be affected by changes to
the D-Bus system bus daemon.
The disadvantage to using the private socket however is security, init(8) only permits the root user to communicate over this socket
which means that read-only commands such as status and list cannot be made by other users.
The --system option instructs initctl to communicate via the D-Bus system bus rather than over the private socket.
This is only possible if the system bus daemon is running and if init(8) is connected to it. The advantage is that the default
security configuration allows non-root users to use read-only commands.
--dest Specifies the well-known name of the init(8) daemon when using --system.
There is normally no need to use this option since the init(8) daemon uses the default com.ubuntu.Upstart name. However it may be
useful for debugging.
--no-wait
Applies to the start, stop, restart and emit commands.
Normally initctl will wait for the command to finish before returning.
For the start, stop and restart commands, finishing means that the named job is running (or has finished for tasks) or has been
fully stopped.
For the emit command, finishing means that all of the jobs affected by the event are running (or have finished for tasks) or have
been fully stopped.
This option instead causes these commands to only wait for the goal change or event to be queued.
--quiet
Reduces output of all commands to errors only.
COMMANDS
start JOB [KEY=VALUE]...
Requests that a new instance of the named JOB be started, outputting the status of the job to standard output when the command com-
pletes.
See status for a description of the output format.
The optional KEY=VALUE arguments specify environment variables to be passed to the starting job, and placed in its environment.
They also serve to specify which instance of multi-instance jobs should be started.
Most jobs only permit a single instance; those that use the instance stanza in their configuration define a string expanded from
environment variables to name the instance. As many unique instances may be started as unique names may be generated by the stanza.
Thus the environment variables also serve to select which instance of JOB is to be acted upon.
If the job is already running, start will return an error.
When called from the pre-stop stanza of a job configuration, start may be called without argument to cancel the stop.
stop JOB [KEY=VALUE]...
Requests that an instance of the named JOB be stopped, outputting the status of the job to standard output when the command com-
pletes.
See status for a description of the output format and start for a discussion on instances.
When called from the pre-start stanza of a job configuration, stop may be called without an argument to cancel the start.
restart
JOB [KEY=VALUE]...
Requests that an instance of the named JOB be restarted, outputting the status of the job to standard output when the command com-
pletes.
The job instance being restarted will retain its original configuration. To have the new instance run with the latest job configu-
ration, stop the job and then start it again instead.
See status for a description of the output format and start for a discussion on instances.
Note that this command can only be used when there is an instance of JOB, if there is none then it returns an error instead of
starting a new one.
reload JOB [KEY=VALUE]...
Sends the SIGHUP signal to running process of the named JOB instance.
See start for a discussion on instances.
status JOB [KEY=VALUE]...
Requests the status an instance of the named JOB, outputting to standard output.
See start for a discussion on instances.
For a single-instance job a line like the following is output:
job start/running, process 1234
The job name is given first followed by the current goal and state of the selected instance. The goal is either start or stop, the
status may be one of waiting, starting, pre-start, spawned, post-start, running, pre-stop, stopping, killed or post-stop.
If the job has an active process, the process id will follow on the same line. If the state is pre-start or post-stop this will be
the process id of the equivalent process, otherwise it will be the process id of the main process.
job start/pre-start, process 902
The post-start and pre-stop states may have multiple processes attached, the extra processes will follow on consecutive lines
indented by a tab:
job start/post-start, process 1234
post-start process 1357
If there is no main process, they may follow on the same line but will be prefixed to indicate that it is not the main process id
being given:
job start/post-start, (post-start) process 1357
Jobs that permit multiple instances have names for each instance, the output is otherwise identical to the above except that the
instance name follows the job name in parentheses:
job (tty1) start/post-start, process 1234
post-start process 1357
list
Requests a list of the known jobs and instances, outputs the status of each to standard output.
Note that this command includes in the enumeration as-yet-to-run jobs (in other words configuration files for which no job instances
have yet been created) in the output with status "stop/waiting". In effect such entries denote configuration files which represent
potential future jobs.
See status for a description of the output format and start for a discussion on instances.
No particular order is used for the output, and there is no difference in the output (other than the instance name appearing in
parentheses) between single-instance and multiple-instance jobs.
emit EVENT [KEY=VALUE]...
Requests that the named EVENT be emitted, potentially causing jobs to be started and stopped depending on their use of the start on
and stop on stanzas in their configuration.
The optional KEY=VALUE arguments specify environment variables to be included with the event and thus exported into the environment
of any jobs started and stopped by the event.
The environment may also serve to specify which instance of multi-instance jobs should be started or stopped. See start for a dis-
cussion on instances.
There is no limitation on the event names that may be emitted with this command, you are free to invent new events and use them in
your job configurations.
The most well-known event used by the default Upstart configuration is the runlevel(7) event. This is normally emitted by the
telinit(8) and shutdown(8) tools.
reload-configuration
Requests that the init(8) daemon reloads its configuration.
This command is generally not necessary since init(8) watches its configuration directories with inotify(7) and automatically
reloads in cases of changes.
No jobs will be started by this command.
version
Requests and outputs the version of the running init daemon.
log-priority
[PRIORITY]
When called with a PRIORITY argument, it requests that the init(8) daemon log all messages with that priority or greater. This may
be used to both increase and decrease the volume of logged messages.
PRIORITY may be one of debug, info, message, warn, error or fatal.
When called without argument, it requests the current minimum message priority that the init(8) daemon will log and outputs to stan-
dard output.
show-config
[OPTIONS] [CONF]
Display emits, start on and stop on job configuration details (in that order) for specified job configuration, CONF. If CONF is not
specified, list information for all valid job configurations.
Note that a job configuration is the name of a job configuration file, without the extension. Note too that this information is
static: it does not refer to any running job.
For each event emitted, a separate line is displayed beginning with two space characters followed by, 'emits event' where 'event'
denotes a single emitted event.
The start on and stop on conditions are listed on separate lines beginning with two space characters and followed by 'start on' and
'stop on' respectively and ending with the appropriate condition.
If a job configuration has no emits, start on, or stop on conditions, the name of the job configuration will be displayed with no
further details.
Note that the start on and stop on conditions will be fully bracketed, regardless of whether they appear like this in the job con-
figuration file. This is useful to see how the init(8) daemon perceives the condition.
Example output:
foo
emits boing
emits blip
start on (starting A and (B or C var=2))
stop on (bar HELLO=world testing=123 or stopping wibble)
OPTIONS
-e, --enumerate
If specified, rather than listing the precise start on and stop on conditions, outputs the emits lines along with one line
for each event or job the CONF in question may be started or stopped by if it were to become a job. If the start on condition
specifies a non-job event, this will be listed verbatim, whereas for a job event, the name of the job as opposed to the event
the job emits will be listed.
The type of entity, its triggering event (if appropriate) and its full environment is displayed in brackets following its
name for clarity.
This option is useful for tools which generate graphs of relationships between jobs and events. It is also instructive since
it shows how the init(8) daemon has parsed the job configuration file.
Example output (an analog of the default output format above):
foo
emits boing
emits blip
start on starting (job: A, env:)
start on B (job:, env:)
start on C (job:, env: var=2)
stop on bar (job:, env: HELLO=world testing=123)
stop on stopping (job: wibble, event: stopping, env:)
check-config
[OPTIONS] [CONF]
Considers all job configurations looking for jobs that cannot be started or stopped, given the currently available job configura-
tions. This is achieved by considering the start on, stop on and emits stanzas for each job configuration and identifying unreach-
able scenarios.
This option is useful for determining the impact of adding or removing job configuration files.
Note that to use this command, it is necessary to ensure that all job configuration files advertise the events they emit correctly.
If errors are identified, the name of the job configuration will be displayed. Subsequent lines will show the failed conditions for
the job configuration, one per line. Condition lines begin with two spaces and are followed with either "start on: " or "stop on: ",
the word "unknown", the type of entity that is not known and finally its name.
Note that only job configurations that are logically in error (those with unsatisfiable conditions) will be displayed. Note too that
job configurations that are syntactically invalid may trigger an error if they would cause a condition to be in error.
Assuming job configuration file /etc/init/foo.conf contains the following:
start on starting grape
stop on peach
The check-config command might display:
foo
start on: unknown job grape
stop on: unknown event peach
If any errors are detected, the exit code will be 1 (one). If all checks pass, the exit code will be 0 (zero).
Note that for complex start on and stop on conditions, this command may give what appears to be misleading output when an error con-
dition is found since all expressions in the failing condition that are in error will generate error output. For example, if job
configuration /etc/init/bar.conf contains the following:
start on (A and (started B or (starting C or D)))
And only event A can be satisfied, the output will be:
bar
start on: unknown job B
start on: unknown job C
start on: unknown event D
OPTIONS
-i [EVENTS], --ignore-events [EVENTS]
If specified, the argument should be a list of comma-separated events to ignore when checking the job configuration files.
This option may be useful to ignore errors if a particular job configuration file does not advertise it emits an event.
Note that internal events (such as startup(7) and starting(7)) are automatically ignored.
-w, --warn
If specified, treat any unknown jobs and events as errors.
notify-disk-writeable
Notify the init(8) daemon that the disk is now writeable. This currently causes the init(8) daemon to flush its internal cache of
'early job' output data. An early job is any job which finishes before the log disk becomes writeable. If job logging is not dis-
abled, this command should be called once the log disk becomes writeable to ensure that output from all early jobs is flushed. If
the data is written successfully to disk, the internal cache is deleted.
usage JOB [KEY=VALUE]...
Show usage information an instance of the named JOB defined with usage stanza.
For job with usage stanza a line like the following is output, see init(5) :
Usage: tty DEV=ttyX - where X is console id
AUTHOR
Written by Scott James Remnant <scott@netsplit.com> and James Hunt <james.hunt@canonical.com>.
REPORTING BUGS
Report bugs at <https://launchpad.net/upstart/+bugs>
COPYRIGHT
Copyright (C) 2009-2011 Canonical Ltd.
This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICU-
LAR PURPOSE.
SEE ALSO
init(8) telinit(8) shutdown(8)
Upstart 2011-06-02 initctl(8)