Salt execution framework#

Types of executions#

Salt commands can be executed in different ways:

  • Remote execution - using the salt command from the Salt master
  • Master execution - using salt-run
  • Local execution - using salt-call initiated on the Salt minion

Each command is just a wrapper around an API client interface. The command to run determines where you are executing the command (Salt master or minion) and where you want the command to run (Salt master or minion).

Calling modules locally on a minion#

Salt modules can be called locally on the minion, bypassing the Salt master’s initial publish of the command on the event bus using the salt-call command.

  • The salt module.functions make up a rich API for system management.
  • These module.functions are all available for execution directly on the minion.

Salt-call is an effective tool for both accessing the Salt functions without needing to initially run through the Salt master, but to also debug operations on a minion. Some complex operations, like configuration management can be more easily debugged if run locally with salt-call.

Using salt-call#

Using salt-call is simple. It is used the same way as the salt command but does not require a target (because it is running the module.function on the local minion).

To view network interfaces using salt-call:

salt-call network.interfaces

The documentation is also available using salt-call:

salt-call sys.doc pkg.install

A module.function alias exists for viewing documentation using salt-call:

salt-call -d service

Viewing local grains#

Grains data can be viewed locally on the minion using salt-call in two different ways:

salt-call grains.items


salt-call -g

You may also view specific grain items the same way:

salt-call grains.get os_familiy

The Salt command line execution#

Salt executes jobs in the remote execution framework in an asynchronous fashion. However, the command line works in synchronous mode so that output is more easily viewed at the completion of a job.

If a command times out or if your shell gets disconnected, the command still runs because it has been published on port 4505.

When a job takes longer than the default 5 seconds or a specific time, the LocalClient will publish another request that goes through the same process to check on the job.

In other words, this is the amount of time before the Salt master will check active jobs on the minions or give up and report a time-out for the missing minion return data. As long as the minion responds to the check job status request, the time will reset for another duration.

Command line execution options#

Salt can execute jobs asynchronously using the --async command line option:

salt -G 'os_family:RedHat' pkg.upgrade --async
salt -G 'os_family:RedHat' pkg.upgrade --async command output

Executed command with job ID: 20190220150112550868

This option will immediately return to the command prompt and print out the assigned Job ID.

In addition to running jobs asynchronously, a job may be submitted synchronously (default mode) but still show the assigned Job ID:

salt -G 'os_family:RedHat' pkg.upgrade -v
salt -G 'os_family:RedHat' pkg.upgrade -v command output

Executed command with job ID: 20190220170719865801

This command will operate in the default synchronous manner, but the Job ID will be printed in the terminal.

All Salt commands (salt, salt-call, and salt-run) can be issued with a logging flag in order to see more detailed output in the terminal:

salt -l debug \*ubuntu status.meminfo
salt -l debug \*ubuntu status.meminfo command output

[DEBUG   ] Reading configuration from /etc/salt/master
[DEBUG   ] Including configuration from '/etc/salt/master.d/reactor.conf'
[DEBUG   ] Reading configuration from /etc/salt/master.d/reactor.conf
[DEBUG   ] Using cached minion ID from /etc/salt/minion_id: 20190218-sosfubuntu
[DEBUG   ] Configuration file path: /etc/salt/master
[DEBUG   ] MasterEvent PUB socket URI: /var/run/salt/master/master_event_pub.ipc
[DEBUG   ] MasterEvent PULL socket URI: /var/run/salt/master/master_event_pull.ipc
[DEBUG   ] Initializing new AsyncZeroMQReqChannel for (u'/etc/salt/pki/master', u'20190218-sosf-master_master', u'tcp://', u'clear')
[DEBUG   ] Connecting the Minion to the Master URI (for the return server): tcp://
[DEBUG   ] Trying to connect to: tcp://
[DEBUG   ] Initializing new IPCClient for path:
[DEBUG   ] LazyLoaded local_cache.get_load
[DEBUG   ] Reading minion list from
[DEBUG   ] get_iter_returns for jid 20190220172435508684 sent to set(['20190218-sosf-ubuntu']) will timeout at 17:24:40.515091
[DEBUG   ] jid 20190220172435508684 return from 20190218-sosf-master
[DEBUG   ] return event: {u'20190218-sosf-ubuntu': {u'jid': u'20190220172435508684', u'retcode': 0, ...

Salt job management#

Salt job management begins by understanding:

  • Job IDs
  • Job Cache

Job IDs#

Every Salt job is assigned a unique Job ID. The Job ID is used to track the individual executions.

Job IDs are represented as a jid and are created on the Salt master for each job and sent down with the command.

The Job IDs are timestamps of when the jobs are started.

A job id looks like this:


This Job ID is for a job started: Feb 20, 2019, at 17:24:35 and 508684 microseconds:

Year: 2019
Month: 02
Day: 20
Hour: 17
Minute: 24
Second: 35
Microsecond: 508684

Job cache#

The job cache is the storage system for all executed jobs.

  • The job cache is located in the cachedir under the directory named jobs.
File: /var/cache/salt/master/jobs
  • This directory is cleaned by the Salt master on a regular basis.

The number of hours that old jobs are kept defaults to 24, but it is configured via the keep_jobs option in the Salt master configuration file.

Keep_jobs: 24

It is recommended to store job data in an external job cache (discussed in a later chapter) if a requirement to keep this data is more than 5 days. This figure is dependent on several factors including:

  • Number of jobs needing to be executed in that time span
  • Number of minions being targeted
  • Master resources (disk space)

Running jobs on the Salt master and managing jobs#

The salt command is typed on the Salt master, but Salt sends jobs for remote execution on minions. The salt-run command sends jobs to Salt to run on the Salt master. Jobs that are to be run on the Salt master by salt-run are called runners.

Runners are a specific type of Salt module intended to execute in the environment of the Salt master. Runners will be discussed in greater detail later in the course, but for now, we will discuss job management which employs the Salt Runner module: jobs. The jobs runner module allows for viewing the Salt master’s job cache.

View running jobs#

Currently running jobs can be viewed via the runner module.function:

salt-run command output

    2019, Feb 20 15:01:12.550868
    2019, Feb 20 15:01:12.550868

View previously run jobs#

Jobs that have been executed in the past keep_jobs window can be easily looked up using the jobs runner. jobs.list_jobs will list information about all previously executed jobs in the keep_jobs window:

salt-run jobs.list_jobs
salt-run jobs.list_jobs command output

    Arguments: []
    Start Time: 2019, Feb 20 10:42:53.056848
    Target: '*'
    Target-type: glob
    - [dmesg]
    Start Time: 2019, Feb 20 10:43:01.355086
    Target: '*ubuntu'
    Target-type: glob

With this data, the details of a specific job can be pulled up using the Job ID:

salt-run jobs.lookup_jid 20190220104253056848

To see the status of a currently active job, add the display_progress=True option:

salt-run jobs.lookup_jid 20190220150112550868 display_progress=True
salt-run jobs.lookup_jid 20190220150112550868 display_progress=True command output

    Querying returner: local_cache

Kill and term jobs#

The saltutil execution module contains Salt functions to terminate Salt jobs:

saltutil.term_job will send a termination signal to a job (SIGTERM 15)

salt 201190218-sosf-redhat saltutil.term)job 20190220150112550868

saltutil.kill_job will send a kill signal to a job (SIGKILL 9)

salt 20190218-sosf-redhat saltutil.kill_job 20190220150112550868

The event system#

Salt maintains an event system that fires local publications on a local UNIX socket.

  • Events are fired for a number of situations on the Salt master.
  • The event system is made available on the Salt master and minion.
  • The same system user that the Salt master or minion is running as can fire events using the Salt event API.

Types of Salt events#

The Salt master has the following types of events:

  • authentication
  • start
  • key
  • job
  • presence
  • cloud
  • run

Event components#

The event system sends two pieces of information. The tag and the data items.

  • The tag is a “/” separated string representing a simple data structure.
  • The data will be serialized by Salt into a MessagePack string.

Viewing Salt events#

One of the best ways to see exactly what events are fired and what data is available in each event is to use the state.event runner:

salt-run state.event pretty=True
salt-run state.event pretty=True command output

salt/job/20190220181913504496/new {
"_stamp": "2019-02-20T18:19:13.506890",
"arg": [],
"fun": "",
"jid": "20190220181913504496",
"minions": [
"missing": [],
"tgt": [
"tgt_type": "glob",
"user": "root"
salt/job/20190220181913504496/ret/20190218-sosf-ubuntu      {
    "_stamp": "2019-02-20T18:19:08.490451",
    "cmd": "_return",
    "fun": "",
    "fun_args": [],
    "id": "20190218-sosf-ubuntu"
    "jid": "20190220181913504496",
    "retcode": 0,
    "retcode": 0,
    "return": true,
    "success": true

Minions firing events to Salt master#

The minions can fire off events on the Salt master via the event execution module. An event can be sent to the Salt master by using the event.send function:

salt-call event.send 'mycustom/app/tag' '{"app": "mycustom", "build_num": "3.1", "result": "true"}' with_grains=True

The function event.fire_master can be used to send events to master as well (without the ability to append grains data):

salt-call event.fire_master '{"app": "mycustom", "build_num": "3.1", "result": "true"}' 'mycustom/app/tag'

Here is some sample output from state.event of an event using event.send and with_grains=True from the Salt minion:

mycustom/app/tag    {
"_stamp": "2019-02-20T18:36:00.182479",
"cmd": "_minion_event",
"data": {
    "__pub_fun": "event.send",
    "__pub_jid": "20190220113600174263",
    "__pub_pid": 12819,
    "__pub_tgt": "salt-call",
    "app": "mycustom",
    "build_num": "3.1",
    "grains": {
        "SSDs": [
    "biosreleasedate": "07/18/2018",
    "biosversion": "1.8.0",
"id": "20190218-sosf-centos",
"tag": "mycustom/app/tag"