Automation wrapper for bash and python commands
automatix is a wrapper for scripted sysadmin tasks. It offers some useful functionality for easier scripting and having full control over the automated process.
The idea of automatix is to write down all the commands you would normally type to your commandline or python console into a YAML file. Then use automatix to execute these commands.
There are different modes for automatix to work. Without any parameters automatix will try to execute the specified command pipeline from the script file until an error occurs or the pipeline is done. The interactive mode (-i) asks for every single commandline step whether to execute, skip or abort. Forced mode (-f) will also proceed if errors occur.
automatix is originally designed for internal //SEIBERT/MEDIA use. It comes therefore with bundlewrap and teamvault support as well as the possibility to use your own logging library.
Beware that this tool cannot substitute the system administrators brain and it needs a responsible handling, since you can do (and destroy) almost everything with it.
Automatix evaluates YAML files and executes defined commands as shell or python commands. There is no check for harmful commands. Be aware that this can cause critical damage to your system.
Please use the interactive mode and doublecheck commands before executing. Usage of automatix is at your own risk!
Automatix requires Python ≥ 3.8.
pip install automatix
You can specify a path to a configuration YAML file via the environment variable AUTOMATIX_CONFIG. Default location is "~/.automatix.cfg.yaml".
# Path to scripts directory
script_dir: ~/automatix_script_files
# Global constants for use in pipeline scripts
constants:
apt_update: 'apt-get -qy update'
apt_upgrade: 'DEBIAN_FRONTEND=noninteractive apt-get -qy -o Dpkg::Options::=--force-confold --no-install-recommends upgrade'
apt_full_upgrade: 'DEBIAN_FRONTEND=noninteractive apt-get -qy -o Dpkg::Options::=--force-confold --no-install-recommends full-upgrade'
# Encoding
encoding: utf-8
# Path for shell imports
import_path: '.'
# SSH Command used for remote connections
ssh_cmd: 'ssh {hostname} sudo '
# Temporary directory on remote machines for shell imports
remote_tmp_dir: 'automatix_tmp'
# Logger
logger: mylogger
# Logging library (has to implement the init_logger method)
logging_lib: mylib.logging
# Bundlewrap support, bundlewrap has to be installed (default: false)
bundlewrap: true
# Teamvault / Secret support, bundlewrap-teamvault has to be installed (default: false)
teamvault: true
automatix [--help|-h] [--systems [SYSTEM1=ADDRESS_OR_NODENAME ...]] [--vars [VAR1=VALUE1 ...]] [--secrets [SECRET1=SECRETID ...]] [--vars-file VARS_FILE_PATH ] [--print-overview|-p] [--jump-to|-j JUMP_TO] [--steps|-s STEPS] [--interactive|-i] [--force|-f] [--debug|-d] [--] scriptfile
scriptfile : The only required parameter for this tool to work. Use " -- " if needed to delimit this from argument fields. See SCRIPTFILE section for more information.
-h, --help : View help message and exit.
--systems SYSTEM1=ADDRESS_OR_NODENAME : Use this to set systems without adding them to the scriptfile or to overwrite them. You can specify multiple systems like: --systems v1=string1 v2=string2 v3=string3
--vars VAR1=VALUE1 : Use this to set vars without adding them to the scriptfile or to overwrite them. You can specify multiple vars like: --vars v1=string1 v2=string2 v3=string3
--secrets SECRET1=SECRETID : Use this to set secrets without adding them to the scriptfile or to overwrite them. You can specify multiple secrets like: --secrets v1=string1 v2=string2 v3=string3 (only if teamvault is enabled)
--vars-file VARS_FILE_PATH
: Use this to specify a CSV file from where automatix reads
systems, variables and secrets. First row must contain the field
types and names. You may also specify an label
field.
Example: label,systems:mysystem,vars:myvar
. The automatix script will
be processed for each row sequentially.
--print-overview, -p : Just print command pipeline overview with indices then exit without executing the commandline. Note that the always pipeline will be executed anyway.
--jump-to JUMP_TO, -j JUMP_TO : Jump to step with index JUMP_TO instead of starting at the beginning. Use -p or the output messages to determine the desired step index. You can use negative numbers to start counting from the end.
--steps STEPS, -s STEPS
: Only execute these steps (comma-separated indices) or exclude steps
by prepending the comma-separated list with "e".
Examples: -s 1,3,7
, -s e2
, -s e0,5,7,2
--interactive, -i : Confirm actions before executing.
--force, -f : Try always to proceed (except manual steps), even if errors occur (no retries).
--debug, -d : Activate debug log level.
automatix -i --systems source=sourcesystem.com target=targetsystem.org -- scriptfile.yaml
The scriptfile describes your automated process. Therefore it contains information about systems, variables, secrets and the command pipeline.
You can provide a path to your scriptfile or place your scriptfile in the predefined directory (see CONFIGURATION section, script_dir). The path has precedence over the predefined directory, if the file exists at both locations.
The scriptfile has to contain valid YAML.
name: Migration Server XY
# Systems you like to refer to in pipeline (accessible via 'SYSTEMS.source')
# If Bundlewrap support is activated use node names instead of hostnames or add preceeding 'hostname!'.
require_version: '1.5.0'
systems:
source: sourcesystem.com
target: targetsystem.org
# Custom vars to use in pipeline
vars:
version: 1.2.3
domain: 'bla.mein-test-system'
# Teamvault Secrets, if activated (left: like vars, right: SECRETID_FIELD, FIELD=username|password|file)
secrets:
web_user: v6GQag_username
web_pw: v6GQag_password
# Imports for functions you like to use (path may be modified in configuration)
imports:
- myfunctions.sh
# like command pipeline but will be exectuted always beforehand
always:
- python: |
import mylib as nc
PERSISTENT_VARS.update(locals())
pipeline:
- remote@target: systemctl stop server
- remote@source: zfs snapshot -r tank@before-migration
- manual: Please trigger preparing tasks via webinterface
- myvar=local: curl -L -vvv -k https://{domain}/
- local: echo "1.1.1.1 {domain}" >> /etc/hosts
- sla=python: NODES.source.metadata.get('sla')
- python: |
sla = '{sla}'
if sla == 'gold':
print('Wow that\'s pretty cool. You have SLA Gold.')
else:
print('Oh. Running out of money? SLA Gold is worth it. You should check your wallet.')
PERSISTENT_VARS['sla'] = sla
- cond=python: sla == 'gold'
- cond?local: echo "This command is only executed if sla is gold."
cleanup:
- local: rm temp_files
name (string) : Just a name for the process. Does not do anything.
require_version (string) : Minimum required Automatix version for this script to run.
systems (associative array)
: Define some systems. Value has to be a valid SSH destination like an
IP address or hostname. If Bundlewrap support is enabled, it has to
be a valid and existing Bundlewrap node or group name, or you can
precede your IP or hostname with hostname!
to define a
non-Bundlewrap system.
You can refer to these systems in the command pipeline in multiple ways:
-
remote@systemname as your command action (see below)
-
via {SYSTEMS.systemname} which will be replaced with the value
-
via NODES.systemname in python actions to use the Bundlewrap node object (Bundlewrap nodes only, no groups)
vars (associative array) : Define some vars. These are accessible in the command pipeline via {varname}. Note: Only valid Python variable names are allowed. You can use "FILE_" prefix followed by a file path to assign the file content to the variable.
secrets (associative array) : Define teamvault secrets. Value has to be in this format: SECRETID_FIELD. FIELD must be one of username, password or file. The resolved secret values are accessible in command line via {secretname}. (only if teamvault is enabled)
imports (list) : Listed shell files (see CONFIGURATION section, import_path) will be sourced before every local or remote command execution. For remote commands, these files are transferred via tar and ssh to your home directory on the remote system beforehand and deleted afterwards. This is meant to define some functions you may need.
always, cleanup (list of associative arrays) : See ALWAYS / CLEANUP PIPELINE section.
pipeline (list of associative arrays) : See PIPELINE section.
Here you define the commands automatix shall execute.
KEY: One of these possible command actions:
-
manual: Some manual instruction for the user. The user has to confirm, that automatix may proceed.
-
local: Local shell command to execute. Imports will be sourced beforehand. /bin/bash will be used for execution.
-
remote@systemname: Remote shell command to execute. Systemname has to be a defined system. The command will be run via SSH (without pseudo-terminal allocation). It uses the standard SSH command. Therefore your .ssh/config should be respected. If systemname is a Bundlewrap group, the remote command will be executed sequentially for every node.
-
python: Python code to execute.
- Notice that there are some modules, constants and functions which
are already imported (check command.py): e.g.
re, subprocess, quote(from shlex)
. The variablevars
is used to store the Automatix variables as a dictionary. You can use it to access or change the variables directly. - If bundlewrap is enabled, the Bundlewrap repository object is
available via AUTOMATIX_BW_REPO and system node objects are
available via NODES.systemname.
Use
AUTOMATIX_BW_REPO.reload()
to reinitialize the Bundlewrap repository from the file system. This can be useful for using newly created nodes (e.g. remote commands).
- Notice that there are some modules, constants and functions which
are already imported (check command.py): e.g.
ASSIGNMENT: For local, remote and python action you
can also define a variable to which the output will be assigned.
To do this prefix the desired variablename and = before the action
key, e.g. myvar=python: NODES.system.hostname
. Be careful when
working with multiline statements. In python the first line is
likely to set the variable. All variables will be converted to
strings when used to build commands in following steps.
CONDITIONS: You can define the command only to be executed if
your condition variable evaluates to "True" in Python. To achieve
this write the variable name followed by a question mark at the very
beginning like cond?python: destroy_system()
. Be aware that all
output from local or remote commands will lead to a non-empty
string which evaluates to "True" in Python, but empty output will
evaluate to "False". Use !?
instead of ?
to invert the condition.
VALUE: Your command. Variables will be replaced with Python format function. Therefore, use curly brackets to refer to variables, systems, secrets and constants.
Constants are available via CONST.KEY, where KEY is the key of your constants in your CONFIGURATION file. There you can define some widely used constants.
In most cases its a good idea to define your command in quotes to avoid parsing errors, but it is not always necessary. Another way is to use '|' to indicate a literal scalar block. There you can even define whole program structures for python (see example).
Because automatix uses Python's format() function:
{
-> {{
}
-> }}
Standard YAML escapes (see also https://yaml.org/spec/1.2/spec.html):
'
-> ''
"
-> \"
\
-> \\
:
-> Please use quotes (double or single).
Same usage as the 'normal' command pipeline, but will be executed every time at start of automatix (always) or at the end (cleanup) even if aborted (a). The commands are executed without --interactive flag, independend of the specified parameters.
Intended use case for always: python imports or informations that are needed afterwards and do not change anything on systems. You want to have these available even if using --jump|-j feature.
Intended use case for cleanup: Remove temporary files or artifacts.
AUTOMATIX_CONFIG: Specify the path to the configuration file. Default is "~/.automatix.cfg.yaml".
AUTOMATIX_SCRIPT_DIR: Set or overwrite directory where script files are located (and searched in case of autocompletion).
AUTOMATIX_TIME: Set this to an arbitrary value to print the times for the single steps and the whole script.
ENCODING: Specify output encoding. Default is "UTF-8".
Additionally you can modify the environment to adjust things to your needs.
If you want to access variables in python action you defined in preceeding command, you can use the PERSISTENT_VARS dictionary (shortcut: PVARS). This is added to the local scope of python actions and the dictonary keys are also available as attributes. Examples:
- To make all local variables of the actual command persistent use
PERSISTENT_VARS.update(locals())
. - To delete one persistent variable named "myvar" use
del PERSISTENT_VARS['myvar']
- To make variable "v2" persistent use
PERSISTENT_VARS['v2'] = v2
orPERSISTENT_VARS.v2 = v2
- Use the shortcut like
PVARS.v2 = v2
You can use variables in PERSISTENT_VARS also as condition by using the shortcut and the attribute notation:
- python: PVARS.cond = some_function()
- PVARS.cond?local: echo 'This is only printed if "some_function" evaluates to "True"'
- PVARS.cond!?local: echo 'And this is printed if "some_function" evaluates to "False"'
An alternative is to make variables global, but in most cases using
PERSISTENT_VARS is more clean. CAUTION: Choosing already existing
(Python) variable names may lead to unexpected behaviour!!! Maybe
you want to check the source code (command.py).
Explanation: automatix is written in Python and uses 'exec' to
execute the command in function context. If you declare variables
globally they remain across commands.
To abort the current automatix and jump to the next batch item you can
raise the SkipBatchItemException
. For aborting the whole automatix
process raise AbortException(return_code: int)
. In both cases the
cleanup pipeline is executed. Same is the case for selecting
a
:abort or c
:continue when asked (interactive or error).
There are different ways to start scripting with automatix. The
author's approach is mainly to consider the process and simply write
down, what to do (manual steps for complex or not automated steps)
and which commands to use.
Then start automatix in interactive mode (-i) and adjust the
single steps one by one. Replace manual steps, if suitable. Whenever
adjustment is needed, abort, adjust and restart automatix with
jump (-j) to the adjusted step.
From automatix 1.13.0 on you can use the reload scriptfile feature
instead. When asked for options (either because a command failed or
you are in interactive mode) you can use -R to reload the
scriptfile. If lines in the scriptfile have changed, or you need to
repeat steps, you can use R+/-$number to reload and adjust the
restart point (available since automatix 1.14.0)
Repeat this procedure to automate more and more and increase quality, whenever you feel like it.
Consider to put often used paths or code sequences in automatix
variables for better readability.
Do the same with variable content like URLs, to make it possible to
overwrite it by command line options. Where ever possible prefer to
use functions to determine already available information, such as BW
metadata, instead of defining things explicitly. This will make
things easier when using the script with different systems /
parameters.
Preferred way of using automatix is to put often used and complex algorithms in shell functions or python libraries (shelllib/pylib) and import them. Advantage of this approach is that you can use your implemented functions multiple times and build up a toolbox of nice functionality over time.
Manual steps will always cause automatix to stop and wait for user input.
Be careful with assignments containing line breaks (echo, ...). Using the variables may lead to unexpected behaviour or errors. From version 1.14.0 on trailing new lines in assignments of Shell commands (local, remote@) are removed.
Assignments containing null bytes are currently not supported.
Because the always pipeline should not change anything, aborting while running this pipeline will not trigger a cleanup.
If you want to abort the pipeline without triggering the cleanup pipeline, use CRTL+C.
While aborting remote functions (via imports), automatix is not able to determine still running processes invoked by the function, because it only checks the processes for the commands (in this case the function name) which is called in the pipeline.
User input questions are of following categories:
- [MS] Manual Step
- [CF] Command Failed
- [PF] Partial command Failed (BW groups)
- [RR] Remote process still Running
- [SE] Syntax Error
Automatix supports bash completion for parameters and the script directory via argcomplete.
Therefor follow the installation instructions for argcomplete, which is at the current time
pip install argcomplete
and either global activation via executing
activate-global-python-argcomplete
or activation for automatix (e.g. in .bashrc
)
eval "$(register-python-argcomplete automatix)"
Automatix will recognize the installed module and offer the completion automatically.