How to make sure an ansible playbook runs completely - bash

Question

I am running this ansible playbook:
---
- hosts: localhost
remote_user: root
tasks:
- name : update system
apt : update_cache=yes
- name : install m4
apt : name=m4 state=present
- name : install build-essential
apt : name=build-essential state=present
- name : install gcc
apt : name=gcc state=present
- name : install gfortran
apt : name=gfortran state=present
- name : install libssl-dev
apt : name=libssl-dev state=present
- name : install python-software-properties
apt : name=python-software-properties state=present
- name : add sage ppa repo
apt_repository: repo='ppa:aims/sagemath'
- name : update system
apt : update_cache=yes
- name : install dvipng
apt : name=dvipng state=present
- name : install sage binary
apt : name=sagemath-upstream-binary state=present
- name : invoke create_sagenb script
command: /usr/bin/screen -d -m sudo /root/databases-and-datamining-iiith/python-scripts/create_sagenb -i -y
- name : invoke start_sage script
command: /usr/bin/screen -d -m sudo /root/databases-and-datamining-iiith/python-scripts/start_sage -i -y
This playbook fails during task "install build-essential" and stops with error asking to run dpkg --configure -a.
How can I make sure that the playbook runs again after facing this error by running the command
dpkg --configure -a
first and then continue with other tasks.

Answer 1

Ansible in general is idempotent. That means you can simply run your playbook again after resolving the issue without conflicts.
This is not always true. In case you have a more complex play and execute tasks depending on the result of another task, this can break easily and a failed task then would bring you into a state that is not so easy to be fixed with Ansible. But that is not the case with the tasks you provided.
If you want to speed things up and skip all the tasks and/or hosts that did not fail, you can work with --limit and/or --start-at-task:
When the playbook fails, you might notice Ansible shows a message including a command which will enable you to limit the play to hosts which failed. So if only 1 host failed you do not need to run the playbook on all hosts:
ansible-playbook ... --limit #/Users/your-username/name-of-playbook.retry
To start at a specific task, you can use --start-at-task. So if your playbook failed at the task "install build-essential" you can start again at right this task and skip all previous tasks:
ansible-playbook ... --start-at-task="install build-essential"
On a side note, the apt module is optimized to work with loops. You can speed up your play by combining the tasks into one single apt task:
tasks:
- name: Install packages that we need for need for apt_repository
apt: update_cache=yes
name={{ item }}
state=present
cache_valid_time=3600
with_items:
- python-software-properties
- python-software-properties-common
- name: add sage ppa repo
apt_repository: repo='ppa:aims/sagemath'
- name: Install packages
apt: update_cache=yes
cache_valid_time=3600
name={{ item }}
state=present
with_items:
- m4
- build-essential
- gcc
- gfortran
- libssl-dev
- dvipng
- sagemath-upstream-binary