Ansible

Личный сайт Go-разработчика из Казани

Содержание

Introduction

1--- 2"{{ Ansible }}" is an orchestration tool written in Python. 3...

Ansible is (one of many) orchestration tools. It allows you to control your environment (infrastructure and code) and automate the manual tasks.

Ansible has great integration with multiple operating systems (even Windows) and some hardware (switches, Firewalls, etc). It has multiple tools that integrate with the cloud providers. Almost every noteworthy cloud provider is present in the ecosystem (AWS, Azure, Google, DigitalOcean, OVH, etc…).

But ansible is way more! It provides execution plans, an API, library, and callbacks.

Main pros and cons

Pros

  • It is an agent-less tool. In most scenarios, it uses ssh as a transport layer. In some way you can use it as ‘bash on steroids’.
  • It is very easy to start. If you are familiar with the concept of ssh - you already know Ansible (ALMOST).
  • It executes ‘as is’ - other tools (salt, puppet, chef - might execute in different scenario than you would expect)
  • Documentation is at the world-class standard!
  • Writing your own modules and extensions is fairly easy.
  • Ansible AWX is the open source version of Ansible Tower we have been waiting for, which provides an excellent UI.

Cons

  • It is an agent-less tool - every agent consumes up to 16MB ram - in some environments, it may be noticeable amount.
  • It is agent-less - you have to verify your environment consistency ‘on-demand’ - there is no built-in mechanism that would warn you about some change automatically (this can be achieved with reasonable effort)
  • Official GUI - Ansible Tower - is great but expensive.
  • There is no ‘small enterprise’ payment plan, however Ansible AWX is the free open source version we were all waiting for.

Neutral

Migration - Ansible <-> Salt is fairly easy - so if you would need an event-driven agent environment - it would be a good choice to start quick with Ansible, and convert to Salt when needed.

Some concepts

Ansible uses ssh or paramiko as a transport layer. In a way you can imagine that you are using a ssh with API to perform your action. The simplest way is to execute remote command in more controlled way (still using ssh). On the other hand - in advanced scope - you can wrap Ansible (use python Ansible code as a library) with your own Python scripts! It would act a bit like Fabric then.

Example

An example playbook to install apache and configure log level

1--- 2- hosts: apache 3 4 vars: 5 apache2_log_level: "warn" 6 7 handlers: 8 - name: restart apache 9 service: 10 name: apache2 11 state: restarted 12 enabled: True 13 notify: 14 - Wait for instances to listen on port 80 15 become: True 16 17 - name: reload apache 18 service: 19 name: apache2 20 state: reloaded 21 notify: 22 - Wait for instances to listen on port 80 23 become: True 24 25 - name: Wait for instances to listen on port 80 26 wait_for: 27 state: started 28 host: localhost 29 port: 80 30 timeout: 15 31 delay: 5 32 33 tasks: 34 - name: Update cache 35 apt: 36 update_cache: yes 37 cache_valid_time: 7200 38 become: True 39 40 - name: Install packages 41 apt: 42 name={{ item }} 43 with_items: 44 - apache2 45 - logrotate 46 notify: 47 - restart apache 48 become: True 49 50 - name: Configure apache2 log level 51 lineinfile: 52 dest: /etc/apache2/apache2.conf 53 line: "LogLevel {{ apache2_log_level }}" 54 regexp: "^LogLevel" 55 notify: 56 - reload apache 57 become: True 58...

Installation

1# Universal way 2$ pip install ansible 3 4# Debian, Ubuntu 5$ apt-get install ansible

Your first ansible command (shell execution)

1# Command pings localhost (defined in default inventory: /etc/ansible/hosts) 2$ ansible -m ping localhost 3# You should see this output 4localhost | SUCCESS => { 5 "changed": false, 6 "ping": "pong" 7}

Shell Commands

There are few commands you should know about

  • ansible (to run modules in CLI)
  • ansible-playbook (to run playbooks)
  • ansible-vault (to manage secrets)
  • ansible-galaxy (to install roles from github/galaxy)

Module

A program (usually python) that executes, does some work and returns proper JSON output. This program performs specialized task/action (like manage instances in the cloud, execute shell command). The simplest module is called ping - it just returns a JSON with pong message.

Example of modules:

  • Module: ping - the simplest module that is useful to verify host connectivity
  • Module: shell - a module that executes a shell command on a specified host(s).
1$ ansible -m ping all 2$ ansible -m shell -a 'date; whoami' localhost #hostname_or_a_group_name
  • Module: command - executes a single command that will not be processed through the shell, so variables like $HOME or operands like |` `; will not work. The command module is more secure, because it will not be affected by the user’s environment. For more complex commands - use shell module.
1$ ansible -m command -a 'date; whoami' # FAILURE 2$ ansible -m command -a 'date' all 3$ ansible -m command -a 'whoami' all
  • Module: file - performs file operations (stat, link, dir, …)
  • Module: raw - executes a low-down and dirty SSH command, not going through the module subsystem (useful to install python2.7)

Task

Execution of a single Ansible module is called a task. The simplest module is called ping as you could see above.

Another example of the module that allows you to execute a command remotely on multiple resources is called shell. See above how you were using them already.

Playbook

Execution plan written in a form of script file(s) is called playbook. Playbooks consist of multiple elements -

  • a list (or group) of hosts that ’the play’ is executed against
  • task(s) or role(s) that are going to be executed
  • multiple optional settings (like default variables, and way more)

Playbook script language is YAML. You can think that playbook is very advanced CLI script that you are executing.

Example of the playbook

This example-playbook would execute (on all hosts defined in inventory) two tasks:

  • ping that would return message pong
  • shell that execute three commands and return the output to our terminal
1- hosts: all 2 3 tasks: 4 - name: "ping all" 5 ping: 6 7 - name: "execute a shell command" 8 shell: "date; whoami; df -h;"

Run the playbook with the command:

1$ ansible-playbook path/name_of_the_playbook.yml

Note: Example playbook is explained in the next chapter: ‘Roles’

More on ansible concept

Inventory

An inventory is a set of objects or hosts, against which we are executing our playbooks or single tasks via shell commands. For these few minutes, let’s assume that we are using the default ansible inventory (which in Debian based system is placed in /etc/ansible/hosts).

localhost

[some_group]
hostA.mydomain.com
hostB.localdomain
1.2.3.4

[a_group_of_a_groups:children]
some_group
some_other_group

ansible-roles (a ’template-playbooks’ with right structure)

You already know that the tasks (modules) can be run via CLI. You also know the playbooks - the execution plans of multiple tasks (with variables and logic).

A concept called role was introduced for parts of the code (playbooks) that should be reusable.

Role is a structured way to manage your set of tasks, variables, handlers, default settings, and way more (meta, files, templates). Roles allow reusing the same parts of code in multiple playbooks (you can parametrize the role ‘further’ during its execution). Its a great way to introduce object oriented management for your applications.

Role can be included in your playbook (executed via your playbook).

1- hosts: all 2 3 tasks: 4 - name: "ping all" 5 ping: 6 - name: "execute a shell command" 7 shell: "date; whoami; df -h;" 8 9 roles: 10 - some_role 11 - { role: another_role, some_variable: 'learnxiny', tags: ['my_tag'] } 12 13 pre_tasks: 14 - name: some pre-task 15 shell: echo 'this task is the last, but would be executed before roles, and before tasks'

For remaining examples we would use additional repository

This example installs ansible in virtualenv so it is independent from the system. You need to initialize it into your shell-context with the source environment.sh command.

We are going to use this repository with examples: https://github.com/sirkubax/ansible-for-learnXinYminutes

1$ # The following example contains a shell-prompt to indicate the venv and relative path 2$ git clone git@github.com:sirkubax/ansible-for-learnXinYminutes.git 3user@host:~/$ cd ansible-for-learnXinYminutes 4user@host:~/ansible-for-learnXinYminutes$ source environment.sh 5$ 6$ # First lets execute the simple_playbook.yml 7(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/simple_playbook.yml

Run the playbook with roles example

1$ source environment.sh 2$ # Now we would run the above playbook with roles 3(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/simple_role.yml

Role directory structure 

roles/
   some_role/
     defaults/      # contains default variables
     files/         # for static files
     templates/     # for jinja templates
     tasks/         # tasks
     handlers/      # handlers
     vars/          # more variables (higher priority)
     meta/          # meta - package (role) info

Role Handlers

Handlers are tasks that can be triggered (notified) during execution of a playbook, but they execute at the very end of a playbook. It is the best way to restart a service, check if the application port is active (successful deployment criteria), etc.

Get familiar with how you can use roles in the simple_apache_role example

playbooks/roles/simple_apache_role/
├── tasks
│   └── main.yml
└── templates
    └── main.yml

ansible - variables

Ansible is flexible - it has 21 levels of variable precedence. read more For now you should know that CLI variables have the top priority. You should also know, that a nice way to pool some data is a lookup

Lookups

Awesome tool to query data from various sources!!! Awesome! query from:

  • pipe (load shell command output into variable!)
  • file
  • stream
  • etcd
  • password management tools
  • url
1# read playbooks/lookup.yml 2# then run 3(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/lookup.yml

You can use them in CLI too

1ansible -m shell -a 'echo "{{ my_variable }}"' -e 'my_variable="{{ lookup("pipe", "date") }}"' localhost 2ansible -m shell -a 'echo "{{ my_variable }}"' -e 'my_variable="{{ lookup("pipe", "hostname") }}"' all 3 4# Or use in playbook 5 6(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/lookup.yml

Register and Conditional

Register

Another way to dynamically generate the variable content is the register command. Register is also useful to store an output of a task and use its value for executing further tasks.

(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/register_and_when.yml
1--- 2- hosts: localhost 3 tasks: 4 - name: check the system capacity 5 shell: df -h / 6 register: root_size 7 8 - name: debug root_size 9 debug: 10 msg: "{{ root_size }}" 11 12 - name: debug root_size return code 13 debug: 14 msg: "{{ root_size.rc }}" 15 16# when: example 17 18 - name: Print this message when return code of 'check the system capacity' was ok 19 debug: 20 msg: "{{ root_size.rc }}" 21 when: root_size.rc == 0 22...

Conditionals - when:

You can define complex logic with Ansible and Jinja functions. Most common is usage of when:, with some variable (often dynamically generated in previous playbook steps with register or lookup)

1--- 2- hosts: localhost 3 tasks: 4 - name: check the system capacity 5 shell: df -h / 6 when: some_variable in 'a string' 7 roles: 8 - { role: mid_nagios_probe, when: allow_nagios_probes } 9...

ansible - tags, limit

You should know about a way to increase efficiency by this simple functionality

TAGS

You can tag a task, role (and its tasks), include, etc, and then run only the tagged resources

ansible-playbook playbooks/simple_playbook.yml --tags=tagA,tag_other
ansible-playbook playbooks/simple_playbook.yml -t tagA,tag_other

There are special tags:
    always

--skip-tags can be used to exclude a block of code
--list-tags to list available tags

Read more

LIMIT

You can limit an execution of your tasks to defined hosts

ansible-playbook playbooks/simple_playbook.yml --limit localhost

--limit my_hostname
--limit groupname
--limit some_prefix*
--limit hostname:group #JM

Templates

Templates are a powerful way to deliver some (partially) dynamic content. Ansible uses Jinja2 language to describe the template.

Some static content

{{ a_variable }}

{% for item in loop_items %}
    this line item is {{ item }}
{% endfor %}

Jinja may have some limitations, but it is a powerful tool that you might like.

Please examine this simple example that installs apache2 and generates index.html from the template «playbooks/roles/simple_apache_role/templates/index.html»

1$ source environment.sh 2$ # Now we would run the above playbook with roles 3(venv) user@host:~/ansible-for-learnXinYminutes$ ansible-playbook playbooks/simple_role.yml --tags apache2

Jinja2 CLI

You can use the jinja in the CLI too

1ansible -m shell -a 'echo {{ my_variable }}' -e 'my_variable=something, playbook_parameter=twentytwo' localhost

In fact - jinja is used to template parts of the playbooks too

1# check part of this playbook: playbooks/roles/sys_debug/tasks/debug_time.yml 2- local_action: shell date +'%F %T' 3 register: ts 4 become: False 5 changed_when: False 6 7- name: Timestamp 8 debug: msg="{{ ts.stdout }}" 9 when: ts is defined and ts.stdout is defined 10 become: False

Jinja2 filters

Jinja is powerful. It has many built-in useful functions.

# get first item of the list
{{ some_list | first() }}
# if variable is undefined - use default value
{{ some_variable | default('default_value') }}

Read More

ansible-vault

To maintain infrastructure as code you need to store secrets. Ansible provides a way to encrypt confidential files so you can store them in the repository, yet the files are decrypted on-the-fly during ansible execution.

The best way to use it is to store the secret in some secure location, and configure ansible to use them during runtime.

1# Try (this would fail) 2$ ansible-playbook playbooks/vault_example.yml 3 4$ echo some_very_very_long_secret > ~/.ssh/secure_located_file 5 6# in ansible.cfg set the path to your secret file 7$ vi ansible.cfg 8 ansible_vault_password_file = ~/.ssh/secure_located_file 9 10#or use env 11$ export ANSIBLE_VAULT_PASSWORD_FILE=~/.ssh/secure_located_file 12 13$ ansible-playbook playbooks/vault_example.yml 14 15 # encrypt the file 16$ ansible-vault encrypt path/somefile 17 18 # view the file 19$ ansible-vault view path/somefile 20 21 # check the file content: 22$ cat path/somefile 23 24 # decrypt the file 25$ ansible-vault decrypt path/somefile

dynamic inventory

You might like to know, that you can build your inventory dynamically. (For Ansible) inventory is just JSON with proper structure - if you can deliver that to ansible - anything is possible.

You do not need to reinvent the wheel - there are plenty of ready to use inventory scripts for the most popular Cloud providers and a lot of in-house popular usecases.

AWS example

1$ etc/inv/ec2.py --refresh 2$ ansible -m ping all -i etc/inv/ec2.py

Read more

ansible profiling - callback

Playbook execution takes some time. It is OK. First make it run, then you may like to speed things up. Since ansible 2.x there is built-in callback for task execution profiling.

vi ansible.cfg
# set this to:
callback_whitelist = profile_tasks

facts-cache and ansible-cmdb

You can pull some information about your environment from another host. If the information does not change - you may consider using a facts_cache to speed things up.

vi ansible.cfg

# if set to a persistent type (not 'memory', for example 'redis') fact values
# from previous runs in Ansible will be stored.  This may be useful when
# wanting to use, for example, IP information from one group of servers
# without having to talk to them in the same playbook run to get their
# current IP information.
fact_caching = jsonfile
fact_caching_connection = ~/facts_cache
fact_caching_timeout = 86400

I like to use jsonfile as my backend. It allows to use another project ansible-cmdb (project on GitHub) that generates a HTML page of your inventory resources. A nice ‘free’ addition!

Debugging ansible [chapter in progress]

When your job fails - it is good to be effective with debugging.

  1. Increase verbosity by using multiple -v [ -vvvvv]
  2. If variable is undefined - grep -R path_of_your_inventory -e missing_variable
  3. If variable (dictionary or a list) is undefined - grep -R path_of_your_inventory -e missing_variable
  4. Jinja template debug
  5. Strange behaviour - try to run the code ‘at the destination’

Infrastructure as code

You already know, that ansible-vault allows you to store your confidential data along with your code. You can go further - and define your ansible installation and configuration as code. See environment.sh to learn how to install the ansible itself inside a virtualenv that is not attached to your operating system (can be changed by non-privileged user), and as additional benefit - upgrading version of ansible is as easy as installing new version in new virtualenv. What is more, you can have multiple versions of Ansible present at the same time.

1# recreate ansible 2.x venv 2$ rm -rf venv2 3$ source environment2.sh 4 5# execute playbook 6(venv2)$ ansible-playbook playbooks/ansible1.9_playbook.yml # would fail - deprecated syntax 7 8# now lets install ansible 1.9.x next to ansible 2.x 9(venv2)$ deactivate 10$ source environment.1.9.sh 11 12# execute playbook 13(venv1.9)$ ansible-playbook playbooks/ansible1.9_playbook.yml # works! 14 15# please note that you have both venv1.9 and venv2 present - you need to (de)activate one - that is all

become-user, become

In Ansible - to become sudo - use the become parameter. Use become_user to specify the username.

- name: Ensure the httpd service is running
  service:
    name: httpd
    state: started
  become: true

Note: You may like to execute Ansible with --ask-sudo-pass or add the user to sudoers file in order to allow non-supervised execution if you require ‘admin’ privileges.

Read more

Tips and tricks

–check -C

Always make sure that your playbook can execute in ‘dry run’ mode (–check), and its execution is not declaring ‘Changed’ objects.

–diff -D

Diff is useful to see nice detail of the files changed. It compare ‘in memory’ the files like diff -BbruN fileA fileB.

Execute hosts with ‘regex’

1ansible -m ping web*

Host groups can be joined, negated, etc

1ansible -m ping web*:!backend:monitoring:&allow_change

Tagging

You should tag some (not all) objects - a task in a playbook, all tasks included form a role, etc. It allows you to execute the chosen parts of the playbook.

no_logs: True

You may see, that some roles print a lot of output in verbose mode. There is also a debug module. This is the place where credentials may leak. Use no_log to hide the output.

Debug module

allows to print a value to the screen - use it!

Register the output of a task

You can register the output (stdout), rc (return code), stderr of a task with the register command.

Conditionals: when:

Loop: with, with_items, with_dict, with_together

Read more

Additional Resources