Note
Welcome to SaltStack! I am excited that you are interested in Salt and starting down the path to better infrastructure management. I developed (and am continuing to develop) Salt with the goal of making the best software available to manage computers of almost any kind. I hope you enjoy working with Salt and that the software can solve your real world needs!
Salt is a different approach to infrastructure management, founded on the idea that high-speed communication with large numbers of systems can open up new capabilities. This approach makes Salt a powerful multitasking system that can solve many specific problems in an infrastructure.
The backbone of Salt is the remote execution engine, which creates a high-speed,
secure and bi-directional communication net for groups of systems. On top of this
communication system, Salt provides an extremely fast, flexible and easy-to-use
configuration management system called Salt States
.
SaltStack has been made to be very easy to install and get started. Setting up Salt should be as easy as installing Salt via distribution packages on Linux or via the Windows installer. The installation documents cover platform-specific installation in depth.
Salt functions on a master/minion topology. A master server acts as a
central control bus for the clients, which are called minions
. The minions
connect back to the master.
Turning on the Salt Master is easy -- just turn it on! The default configuration is suitable for the vast majority of installations. The Salt Master can be controlled by the local Linux/Unix service manager:
On Systemd based platforms (OpenSuse, Fedora):
systemctl start salt-master
On Upstart based systems (Ubuntu, older Fedora/RHEL):
service salt-master start
On SysV Init systems (Debian, Gentoo etc.):
/etc/init.d/salt-master start
Alternatively, the Master can be started directly on the command-line:
salt-master -d
The Salt Master can also be started in the foreground in debug mode, thus greatly increasing the command output:
salt-master -l debug
The Salt Master needs to bind to two TCP network ports on the system. These ports
are 4505
and 4506
. For more in depth information on firewalling these ports,
the firewall tutorial is available here.
Note
The Salt Minion can operate with or without a Salt Master. This walk-through assumes that the minion will be connected to the master, for information on how to run a master-less minion please see the master-less quick-start guide:
The Salt Minion only needs to be aware of one piece of information to run, the network location of the master.
By default the minion will look for the DNS name salt
for the master,
making the easiest approach to set internal DNS to resolve the name salt
back to the Salt Master IP.
Otherwise, the minion configuration file will need to be edited so that the
configuration option master
points to the DNS name or the IP of the Salt Master:
Note
The default location of the configuration files is /etc/salt
. Most
platforms adhere to this convention, but platforms such as FreeBSD and
Microsoft Windows place this file in different locations.
/etc/salt/minion:
master: saltmaster.example.com
Now that the master can be found, start the minion in the same way as the master; with the platform init system or via the command line directly:
As a daemon:
salt-minion -d
In the foreground in debug mode:
salt-minion -l debug
When the minion is started, it will generate an id
value, unless it has
been generated on a previous run and cached in the configuration directory, which
is /etc/salt
by default. This is the name by which the minion will attempt
to authenticate to the master. The following steps are attempted, in order to
try to find a value that is not localhost
:
socket.getfqdn()
is run/etc/hostname
is checked (non-Windows only)/etc/hosts
(%WINDIR%\system32\drivers\etc\hosts
on Windows hosts) is
checked for hostnames that map to anything within 127.0.0.0/8.If none of the above are able to produce an id which is not localhost
, then
a sorted list of IP addresses on the minion (excluding any within
127.0.0.0/8) is inspected. The first publicly-routable IP address is
used, if there is one. Otherwise, the first privately-routable IP address is
used.
If all else fails, then localhost
is used as a fallback.
Note
Overriding the id
The minion id can be manually specified using the id
parameter in the minion config file. If this configuration value is
specified, it will override all other sources for the id
.
Now that the minion is started, it will generate cryptographic keys and attempt to connect to the master. The next step is to venture back to the master server and accept the new minion's public key.
Salt authenticates minions using public-key encryption and authentication. For a minion to start accepting commands from the master, the minion keys need to be accepted by the master.
The salt-key
command is used to manage all of the keys on the
master. To list the keys that are on the master:
salt-key -L
The keys that have been rejected, accepted and pending acceptance are listed. The easiest way to accept the minion key is to accept all pending keys:
salt-key -A
Note
Keys should be verified! Print the master key fingerprint by running salt-key -F master
on the Salt master. Copy the master.pub
fingerprint from the Local Keys section,
and then set this value as the master_finger
in the minion configuration
file. Restart the Salt minion.
On the minion, run salt-key -f minion-id
to print the fingerprint of the
minion's public key. This fingerprint can then be compared against the fingerprint
generated on the minion.
On the master:
# salt-key -f foo.domain.com
Unaccepted Keys:
foo.domain.com: 39:f9:e4:8a:aa:74:8d:52:1a:ec:92:03:82:09:c8:f9
On the minion:
# salt-call key.finger --local
local:
39:f9:e4:8a:aa:74:8d:52:1a:ec:92:03:82:09:c8:f9
If they match, approve the key with salt-key -a foo.domain.com
.
Now that the minion is connected to the master and authenticated, the master can start to command the minion.
Salt commands allow for a vast set of functions to be executed and for specific minions and groups of minions to be targeted for execution.
The salt
command is comprised of command options, target specification,
the function to execute, and arguments to the function.
A simple command to start with looks like this:
salt '*' test.ping
The *
is the target, which specifies all minions.
test.ping
tells the minion to run the test.ping
function.
In the case of test.ping
, test
refers to a execution module. ping
refers to the ping
function contained in the aforementioned test
module.
Note
Execution modules are the workhorses of Salt. They do the work on the system to perform various tasks, such as manipulating files and restarting services.
The result of running this command will be the master instructing all of the
minions to execute test.ping
in parallel
and return the result.
This is not an actual ICMP ping, but rather a simple function which returns True
.
Using test.ping
is a good way of confirming that a minion is
connected.
Note
Each minion registers itself with a unique minion ID. This ID defaults to
the minion's hostname, but can be explicitly defined in the minion config as
well by using the id
parameter.
Of course, there are hundreds of other modules that can be called just as
test.ping
can. For example, the following would return disk usage on all
targeted minions:
salt '*' disk.percent
Salt comes with a vast library of functions available for execution, and Salt
functions are self-documenting. To see what functions are available on the
minions execute the sys.doc
function:
salt '*' sys.doc
This will display a very large list of available functions and documentation on them.
Note
Module documentation is also available on the web.
These functions cover everything from shelling out to package management to manipulating database servers. They comprise a powerful system management API which is the backbone to Salt configuration management and many other aspects of Salt.
Note
Salt comes with many plugin systems. The functions that are available via
the salt
command are called Execution Modules.
The cmd module contains
functions to shell out on minions, such as cmd.run
and cmd.run_all
:
salt '*' cmd.run 'ls -l /etc'
The pkg
functions automatically map local system package managers to the
same salt functions. This means that pkg.install
will install packages via
yum
on Red Hat based systems, apt
on Debian systems, etc.:
salt '*' pkg.install vim
Note
Some custom Linux spins and derivatives of other distributions are not properly
detected by Salt. If the above command returns an error message saying that
pkg.install
is not available, then you may need to override the pkg
provider. This process is explained here.
The network.interfaces
function will
list all interfaces on a minion, along with their IP addresses, netmasks, MAC
addresses, etc:
salt '*' network.interfaces
The default output format used for most Salt commands is called the nested
outputter, but there are several other outputters that can be used to change
the way the output is displayed. For instance, the pprint
outputter can be
used to display the return data using Python's pprint
module:
root@saltmaster:~# salt myminion grains.item pythonpath --out=pprint
{'myminion': {'pythonpath': ['/usr/lib64/python2.7',
'/usr/lib/python2.7/plat-linux2',
'/usr/lib64/python2.7/lib-tk',
'/usr/lib/python2.7/lib-tk',
'/usr/lib/python2.7/site-packages',
'/usr/lib/python2.7/site-packages/gst-0.10',
'/usr/lib/python2.7/site-packages/gtk-2.0']}}
The full list of Salt outputters, as well as example output, can be found here.
salt-call
¶The examples so far have described running commands from the Master using the
salt
command, but when troubleshooting it can be more beneficial to login
to the minion directly and use salt-call
.
Doing so allows you to see the minion log messages specific to the command you
are running (which are not part of the return data you see when running the
command from the Master using salt
), making it unnecessary to tail the
minion log. More information on salt-call
and how to use it can be found
here.
Salt uses a system called Grains to build up static data about minions. This data includes information about the operating system that is running, CPU architecture and much more. The grains system is used throughout Salt to deliver platform data to many components and to users.
Grains can also be statically set, this makes it easy to assign values to minions for grouping and managing.
A common practice is to assign grains to minions to specify what the role or
roles a minion might be. These static grains can be set in the minion
configuration file or via the grains.setval
function.
Salt allows for minions to be targeted based on a wide range of criteria. The
default targeting system uses globular expressions to match minions, hence if
there are minions named larry1
, larry2
, curly1
and curly2
, a
glob of larry*
will match larry1
and larry2
, and a glob of *1
will match larry1
and curly1
.
Many other targeting systems can be used other than globs, these systems include:
The concepts of targets are used on the command line with Salt, but also function in many other areas as well, including the state system and the systems used for ACLs and user permissions.
Many of the functions available accept arguments which can be passed in on the command line:
salt '*' pkg.install vim
This example passes the argument vim
to the pkg.install function. Since
many functions can accept more complex input then just a string, the arguments
are parsed through YAML, allowing for more complex data to be sent on the
command line:
salt '*' test.echo 'foo: bar'
In this case Salt translates the string 'foo: bar' into the dictionary "{'foo': 'bar'}"
Note
Any line that contains a newline will not be parsed by YAML.
Now that the basics are covered the time has come to evaluate States
. Salt
States
, or the State System
is the component of Salt made for
configuration management.
The state system is already available with a basic Salt setup, no additional configuration is required. States can be set up immediately.
Note
Before diving into the state system, a brief overview of how states are constructed will make many of the concepts clearer. Salt states are based on data modeling and build on a low level data structure that is used to execute each state function. Then more logical layers are built on top of each other.
The high layers of the state system which this tutorial will cover consists of everything that needs to be known to use states, the two high layers covered here are the sls layer and the highest layer highstate.
Understanding the layers of data management in the State System will help with understanding states, but they never need to be used. Just as understanding how a compiler functions assists when learning a programming language, understanding what is going on under the hood of a configuration management system will also prove to be a valuable asset.
The state system is built on SLS formulas. These formulas are built out in files on Salt's file server. To make a very basic SLS formula open up a file under /srv/salt named vim.sls. The following state ensures that vim is installed on a system to which that state has been applied.
/srv/salt/vim.sls:
vim:
pkg.installed
Now install vim on the minions by calling the SLS directly:
salt '*' state.sls vim
This command will invoke the state system and run the vim
SLS.
Now, to beef up the vim SLS formula, a vimrc
can be added:
/srv/salt/vim.sls:
vim:
pkg.installed: []
/etc/vimrc:
file.managed:
- source: salt://vimrc
- mode: 644
- user: root
- group: root
Now the desired vimrc
needs to be copied into the Salt file server to
/srv/salt/vimrc
. In Salt, everything is a file, so no path redirection needs
to be accounted for. The vimrc
file is placed right next to the vim.sls
file.
The same command as above can be executed to all the vim SLS formulas and now
include managing the file.
Note
Salt does not need to be restarted/reloaded or have the master manipulated in any way when changing SLS formulas. They are instantly available.
Obviously maintaining SLS formulas right in a single directory at the root of the file server will not scale out to reasonably sized deployments. This is why more depth is required. Start by making an nginx formula a better way, make an nginx subdirectory and add an init.sls file:
/srv/salt/nginx/init.sls:
nginx:
pkg.installed: []
service.running:
- require:
- pkg: nginx
A few concepts are introduced in this SLS formula.
First is the service statement which ensures that the nginx
service is running.
Of course, the nginx service can't be started unless the package is installed --
hence the require
statement which sets up a dependency between the two.
The require
statement makes sure that the required component is executed before
and that it results in success.
Note
The require option belongs to a family of options called requisites. Requisites are a powerful component of Salt States, for more information on how requisites work and what is available see: Requisites
Also evaluation ordering is available in Salt as well: Ordering States
This new sls formula has a special name -- init.sls
. When an SLS formula is
named init.sls
it inherits the name of the directory path that contains it.
This formula can be referenced via the following command:
salt '*' state.sls nginx
Note
Reminder!
Just as one could call the test.ping
or disk.usage
execution modules,
state.sls
is simply another execution module. It simply takes the name of an
SLS file as an argument.
Now that subdirectories can be used, the vim.sls
formula can be cleaned up.
To make things more flexible, move the vim.sls
and vimrc into a new subdirectory
called edit
and change the vim.sls
file to reflect the change:
/srv/salt/edit/vim.sls:
vim:
pkg.installed
/etc/vimrc:
file.managed:
- source: salt://edit/vimrc
- mode: 644
- user: root
- group: root
Only the source path to the vimrc file has changed. Now the formula is
referenced as edit.vim
because it resides in the edit subdirectory.
Now the edit subdirectory can contain formulas for emacs, nano, joe or any other
editor that may need to be deployed.
Two walk-throughs are specifically recommended at this point. First, a deeper run through States, followed by an explanation of Pillar.
An understanding of Pillar is extremely helpful in using States.
Two more in-depth States tutorials exist, which delve much more deeply into States functionality.
These tutorials include much more in-depth information including templating SLS formulas etc.
This concludes the initial Salt walk-through, but there are many more things still to learn! These documents will cover important core aspects of Salt:
A few more tutorials are also available:
This still is only scratching the surface, many components such as the reactor and event systems, extending Salt, modular components and more are not covered here. For an overview of all Salt features and documentation, look at the Table of Contents.