Delegation, Rolling Updates, and Local Actions

• Rolling Update Batch Size
• Maximum Failure Percentage
• Delegation
• Delegated facts
• Run Once
• Local Playbooks
• Interrupt execution on any error

Being designed for multi-tier deployments since the beginning, Ansible is great at doing things on one host on behalf of another, or doing local steps with reference to some remote hosts.

This in particular is very applicable when setting up continuous deployment infrastructure or zero downtime rolling updates, where you might be talking with load balancers or monitoring systems.

Additional features allow for tuning the orders in which things complete, and assigning a batch window size for how many machines to process at once during a rolling update.

This section covers all of these features. For examples of these items in use, please see the ansible-examples repository. There are quite a few examples of zero-downtime update procedures for different kinds of applications.

You should also consult the Working With Modules section, various modules like ‘ec2_elb’, ‘nagios’, and ‘bigip_pool’, and ‘netscaler’ dovetail neatly with the concepts mentioned here.

You’ll also want to read up on Roles, as the ‘pre_task’ and ‘post_task’ concepts are the places where you would typically call these modules.

Be aware that certain tasks are impossible to delegate, i.e. include, add_host, debug, etc as they always execute on the controller.

Rolling Update Batch Size

By default, Ansible will try to manage all of the machines referenced in a play in parallel. For a rolling update use case, you can define how many hosts Ansible should manage at a single time by using the serial keyword:

- name: test play
  hosts: webservers
  serial: 2
  gather_facts: False
  tasks:
  - name: task one
    command: hostname
  - name: task two
    command: hostname

In the above example, if we had 4 hosts in the group ‘webservers’, 2 would complete the play completely before moving on to the next 2 hosts:

PLAY [webservers] ****************************************

TASK [task one] ******************************************
changed: [web2]
changed: [web1]

TASK [task two] ******************************************
changed: [web1]
changed: [web2]

PLAY [webservers] ****************************************

TASK [task one] ******************************************
changed: [web3]
changed: [web4]

TASK [task two] ******************************************
changed: [web3]
changed: [web4]

PLAY RECAP ***********************************************
web1      : ok=2    changed=2    unreachable=0    failed=0
web2      : ok=2    changed=2    unreachable=0    failed=0
web3      : ok=2    changed=2    unreachable=0    failed=0
web4      : ok=2    changed=2    unreachable=0    failed=0

The serial keyword can also be specified as a percentage, which will be applied to the total number of hosts in a play, in order to determine the number of hosts per pass:

- name: test play
  hosts: webservers
  serial: "30%"

If the number of hosts does not divide equally into the number of passes, the final pass will contain the remainder.

As of Ansible 2.2, the batch sizes can be specified as a list, as follows:

- name: test play
  hosts: webservers
  serial:
  - 1
  - 5
  - 10

In the above example, the first batch would contain a single host, the next would contain 5 hosts, and (if there are any hosts left), every following batch would contain 10 hosts until all available hosts are used.

It is also possible to list multiple batch sizes as percentages:

- name: test play
  hosts: webservers
  serial:
  - "10%"
  - "20%"
  - "100%"

You can also mix and match the values:

- name: test play
  hosts: webservers
  serial:
  - 1
  - 5
  - "20%"

Maximum Failure Percentage

By default, Ansible will continue executing actions as long as there are hosts in the batch that have not yet failed. The batch size for a play is determined by the serial parameter. If serial is not set, then batch size is all the hosts specified in the hosts: field. In some situations, such as with the rolling updates described above, it may be desirable to abort the play when a certain threshold of failures have been reached. To achieve this, you can set a maximum failure percentage on a play as follows:

- hosts: webservers
  max_fail_percentage: 30
  serial: 10

In the above example, if more than 3 of the 10 servers in the group were to fail, the rest of the play would be aborted.

Delegation

This isn’t actually rolling update specific but comes up frequently in those cases.

If you want to perform a task on one host with reference to other hosts, use the ‘delegate_to’ keyword on a task. This is ideal for placing nodes in a load balanced pool, or removing them. It is also very useful for controlling outage windows. Be aware that it does not make sense to delegate all tasks, debug, add_host, include, etc always get executed on the controller. Using this with the ‘serial’ keyword to control the number of hosts executing at one time is also a good idea:

---

- hosts: webservers
  serial: 5

  tasks:

  - name: take out of load balancer pool
    command: /usr/bin/take_out_of_pool {{ inventory_hostname }}
    delegate_to: 127.0.0.1

  - name: actual steps would go here
    yum:
      name: acme-web-stack
      state: latest

  - name: add back to load balancer pool
    command: /usr/bin/add_back_to_pool {{ inventory_hostname }}
    delegate_to: 127.0.0.1

These commands will run on 127.0.0.1, which is the machine running Ansible. There is also a shorthand syntax that you can use on a per-task basis: ‘local_action’. Here is the same playbook as above, but using the shorthand syntax for delegating to 127.0.0.1:

---

# ...

  tasks:

  - name: take out of load balancer pool
    local_action: command /usr/bin/take_out_of_pool {{ inventory_hostname }}

# ...

  - name: add back to load balancer pool
    local_action: command /usr/bin/add_back_to_pool {{ inventory_hostname }}

A common pattern is to use a local action to call ‘rsync’ to recursively copy files to the managed servers. Here is an example:

---
# ...
  tasks:

  - name: recursively copy files from management server to target
    local_action: command rsync -a /path/to/files {{ inventory_hostname }}:/path/to/target/

Note that you must have passphrase-less SSH keys or an ssh-agent configured for this to work, otherwise rsync will need to ask for a passphrase.

In case you have to specify more arguments you can use the following syntax:

---
# ...
  tasks:

  - name: Send summary mail
    local_action:
      module: mail
      subject: "Summary Mail"
      to: "{{ mail_recipient }}"
      body: "{{ mail_body }}"
    run_once: True

The ansible_host variable (ansible_ssh_host in 1.x or specific to ssh/paramiko plugins) reflects the host a task is delegated to.

Delegated facts

By default, any fact gathered by a delegated task are assigned to the inventory_hostname (the current host) instead of the host which actually produced the facts (the delegated to host). The directive delegate_facts may be set to True to assign the task’s gathered facts to the delegated host instead of the current one.:

- hosts: app_servers
  tasks:
    - name: gather facts from db servers
      setup:
      delegate_to: "{{item}}"
      delegate_facts: True
      loop: "{{groups['dbservers']}}"

The above will gather facts for the machines in the dbservers group and assign the facts to those machines and not to app_servers. This way you can lookup hostvars[‘dbhost1’][‘ansible_default_ipv4’][‘address’] even though dbservers were not part of the play, or left out by using –limit.

Run Once

In some cases there may be a need to only run a task one time for a batch of hosts. This can be achieved by configuring “run_once” on a task:

---
# ...

  tasks:

    # ...

    - command: /opt/application/upgrade_db.py
      run_once: true

    # ...

This directive forces the task to attempt execution on the first host in the current batch and then applies all results and facts to all the hosts in the same batch.

This approach is similar to applying a conditional to a task such as:

- command: /opt/application/upgrade_db.py
  when: inventory_hostname == webservers[0]

But the results are applied to all the hosts.

Like most tasks, this can be optionally paired with “delegate_to” to specify an individual host to execute on:

- command: /opt/application/upgrade_db.py
  run_once: true
  delegate_to: web01.example.org

As always with delegation, the action will be executed on the delegated host, but the information is still that of the original host in the task.

Local Playbooks

It may be useful to use a playbook locally, rather than by connecting over SSH. This can be useful for assuring the configuration of a system by putting a playbook in a crontab. This may also be used to run a playbook inside an OS installer, such as an Anaconda kickstart.

To run an entire playbook locally, just set the “hosts:” line to “hosts: 127.0.0.1” and then run the playbook like so:

ansible-playbook playbook.yml --connection=local

Alternatively, a local connection can be used in a single playbook play, even if other plays in the playbook use the default remote connection type:

- hosts: 127.0.0.1
  connection: local

Interrupt execution on any error

With the ‘’any_errors_fatal’’ option, any failure on any host in a multi-host play will be treated as fatal and Ansible will exit immediately without waiting for the other hosts.

Sometimes ‘’serial’’ execution is unsuitable; the number of hosts is unpredictable (because of dynamic inventory) and speed is crucial (simultaneous execution is required), but all tasks must be 100% successful to continue playbook execution.

For example, consider a service located in many datacenters with some load balancers to pass traffic from users to the service. There is a deploy playbook to upgrade service deb-packages. The playbook has the stages:

• disable traffic on load balancers (must be turned off simultaneously)
• gracefully stop the service
• upgrade software (this step includes tests and starting the service)
• enable traffic on the load balancers (which should be turned on simultaneously)

The service can’t be stopped with “alive” load balancers; they must be disabled first. Because of this, the second stage can’t be played if any server failed in the first stage.

For datacenter “A”, the playbook can be written this way:

---
- hosts: load_balancers_dc_a
  any_errors_fatal: True
  tasks:
  - name: 'shutting down datacenter [ A ]'
    command: /usr/bin/disable-dc

- hosts: frontends_dc_a
  tasks:
  - name: 'stopping service'
    command: /usr/bin/stop-software
  - name: 'updating software'
    command: /usr/bin/upgrade-software

- hosts: load_balancers_dc_a
  tasks:
  - name: 'Starting datacenter [ A ]'
    command: /usr/bin/enable-dc

In this example Ansible will start the software upgrade on the front ends only if all of the load balancers are successfully disabled.