Why does my Ansible task hang? - ansible

Question

I have the following ansible playbook:
- hosts: node1
sudo: yes
gather_facts: no
tasks:
- name: update apt
apt: update_cache=yes
- name: install python-setuptools
apt: name=python-setuptools update_cache=yes
- name: easy_install pexpect module
easy_install: name=pexpect state=latest
- name: add geonode repo
apt_repository: repo='ppa:geonode/stable' state=present
- name: update apt
apt: update_cache=yes
- name: install geonode
apt: name=geonode update_cache=yes
- expect:
command: geonode createsuperuser
responses:
(?i)username: 'test'
(?i)email: 'test#test.com'
When I run it I get:
PLAY [node1] *******************************************************************
TASK [update apt] **************************************************************
ok: [node1]
TASK [install python-setuptools] ***********************************************
changed: [node1]
TASK [easy_install pexpect module] *********************************************
changed: [node1]
TASK [add geonode repo] ********************************************************
changed: [node1]
TASK [update apt] **************************************************************
ok: [node1]
TASK [install geonode] *********************************************************
Then it hangs indefinitely.
In the remote node (node1), I checked the dir
/home/vagrant/.ansible/tmp/ansible-tmp-1470059145.13-122191240803512/
run the file inside to see why my task is hanging
vagrant#node1:~/.ansible/tmp/ansible-tmp-1470059145.13-122191240803512$ python apt
and get:
{"msg": "Failed to lock apt for exclusive operation", "failed": true, "invocation": {"module_args": {"dpkg_options": "force-confdef,force-confold", "autoremove": false, "force": false, "name": "geonode", "install_recommends": null, "package": ["geonode"], "purge": false, "allow_unauthenticated": false, "state": "present", "upgrade": null, "update_cache": true, "default_release": null, "only_upgrade": false, "deb": null, "cache_valid_time": null}}}
Do you have any insights?
EDIT 1:
It is all day I'm launching this script and never got it working. As I posted this question, obviously, the script successfully executed till the end in 15 minutes. I launched it before lunch today and after 1 hour it was still hanging. Why do I get such a different behaviour? Is there a way in which I can control it?

Answer 1

This issue might be caused by an empty /var/lib/apt folder.
Vagrant might take a while to populate these folders which could cause the apt lock.
Also the playbook is inefficient as update_cache is used multiple times. I would recommend to use something like this:
- hosts: node1
sudo: yes
gather_facts: no
tasks:
# Pause for 5 minutes to make sure vagrant does not hold apt lock.
- pause:
minutes: 5
- name: add geonode repo
apt_repository:
repo: 'ppa:geonode/stable'
state: present
- name: Install apt packages.
apt:
name: "{{ item }}"
state: present
update_cache: true
with_items:
- python-setuptools
- geonode
- name: Create geonode superuser.
expect:
command: geonode createsuperuser
responses:
(?i)username: 'test'
(?i)email: 'test#test.com'
This way Ansible won't update the repositories multiple times during the play.

Answer 2

Since the last thing you see is TASK [install geonode], that's where it's getting stuck.
You are asking it to run geonode createsuperuser which you expect to cause a prompt for username and password.
But what is likely happening is that there is an error being produced from that command, and the expect task is not handling the error, instead just hanging.
You can login to the server you are running this against and run the geonode createsuperuser command manually to see what error is being produced.
In my case, this was happening as a result of the username already being taken since I successfully ran the command on this machine already.
Error: That username is already taken.
Even with the echo: yes argument, ansible doesn't seem to pass on the responses to make it obvious what's happening. And it doesn't accept ignore_errors, so there seems to be no way to handle errors with the expect module.
To work around this, I added another task after the createsuperuser task which places a file in the project indicating the user was created once, and then added creates: {{ path }}/superuser_exists.txt to the createsuperuser task so that it won't run if that file is already there.
It's a hack, but an easy one, and until the module gets better error handling, it will work well enough.
- name: Create the django superuser
expect:
command: "{{ virtualenv_path }}/bin/python3 {{ project_path }}/{{ api_app_name }}/manage.py createsuperuser"
creates: "{{ project_path }}/{{ api_app_name }}/superuser_exists.txt"
responses:
(?i)username: "{{ superuser_username }}"
(?i)email: "{{ superuser_email }}"
(?i)password: "{{ superuser_password }}"
(?i)again: "{{ superuser_password }}"
- name: Create a file to indicate that the superuser was already created
file: path="{{ project_path }}/{{ api_app_name }}/superuser_exists.txt" state=touch