Ansible: install multiple Python packages on a single session

Question

One of my playbooks contains a task that installs basic Python packages:

---
  -
    name: "Install Python packages: {{ python_packages_to_install }}"
    sudo: true
    pip: name={{ item }}
    with_items: python_packages_to_install

With the following list of packages:

-
  include: python_basics.yaml
  vars:
     python_packages_to_install:
       - virtualenv
       - pss
       - requests
       - comment-builder
       - boto
       - ansible
       - uwsgitop
       - gitpull
       - ipython

The task works correctly and installs the packages:

TASK: [common | Install Python packages: ['virtualenv', 'pss', 'requests', 'comment-builder', 'boto', 'ansible', 'uwsgitop', 'gitpull', 'ipython']] ***
ok: [push-prod-01] => (item=virtualenv)
ok: [push-prod-01] => (item=pss)
ok: [push-prod-01] => (item=requests)
ok: [push-prod-01] => (item=comment-builder)
ok: [push-prod-01] => (item=boto)
ok: [push-prod-01] => (item=ansible)
ok: [push-prod-01] => (item=uwsgitop)
ok: [push-prod-01] => (item=gitpull)
changed: [push-prod-01] => (item=ipython)

The problem is that each line is executed using a consecutive SSH command, instead of installing all the packages in a single call.

Is there a way to install multiple Python packages on an Ansible pip command?

Answer 1

Expanding on Ben's answer, you can also continue to preserve the package list as a yaml list like you have it, and do the projection to a single value when you pass it to the pip module like:

pip: name="{{ python_packages_to_install | join(' ') }}"

Keeps your playbook a little more maintainable...

Answer 2

Unlike Ansible apt module, the pip module does not accept a comma-delimited list of packages. Instead, you can provide all the package names as a space delimited string, using the == syntax to specify versions:

python_packages_to_install: "virtualenv==1.11.6 pss requests comment-builder boto ansible uwsgitop gitpull ipython"

If you're like me it may strike you as ugly and impractical to manage. An alternative is to use the requirements option in the pip module to specify a requirements file. Of course you'll need to create the requirements file first, probably using a template.