Sample project having one loadbalancer and two webserver setup using Vagrant, Terraform, Ansible, HA proxy and Testinfra
Following softwares/packages must be install in your system for running this project :-
- Virtualbox (https://www.virtualbox.org/)
- Vagrant (https://www.vagrantup.com/)
- Terraform (https://learn.hashicorp.com/terraform/getting-started/install.html)
- Ansible (https://docs.ansible.com/ansible/latest/installation_guide/intro_installation.html)
- Testinfra (https://testinfra.readthedocs.io/en/latest/)
- Step 1. I have used custom terrform vagrant provider. So first we need to install vagrant plugin for terraform, I have written the script terraform_vagrant_plugin_install.sh for that. We need to run it.
sudo sh scripts/terraform_vagrant_plugin_install.sh
Note: If any issues faced, try alternate setup without terraform.
-
Step 2. RUN
terraform init -
Step 3. RUN
terraform apply -var sticky_session=false --auto-approve. Variable sticky_session is for the deciding the HA proxy routing configuration
sticky_session=false means running with roundrobbin routing policy.
sticky_session=true means running with sticky session routing policy.
-
Step 1. RUN
vagrant up -
Step 2. RUN
ansible-playbook -i inventory/hosts ansible_web_deploy.yaml --extra-vars 'sticky_session=false'. Variable sticky_session is for the deciding the HA proxy routing configuration
sticky_session=false means running with roundrobbin routing policy.
sticky_session=true means running with sticky session routing policy.
Round Robbin Configuration
Repeatedly RUN curl 192.168.33.30 or RUN 192.168.33.30 in browser (192.168.33.30 is loadbalancer ip), we will get alternatively
Stick Session Configuration
Repeatedly RUN curl 192.168.33.30 or RUN 192.168.33.30 in browser (192.168.33.30 is loadbalancer ip), we will get only either one the webpages (based on bind cookie).
If one of the node (webserver) is down the loadbalancer automatically serves traffic from healhy node.
I have used testinfra as a automated testing tool for testing the Infrastructure. More Details (https://testinfra.readthedocs.io/en/latest/).
For testing Webservers
I have written 4 testcases for testing webservers
- TestCase for having root prevelleges
- TestCase for validating apache is installed on the webservers
- TestCase for validating the apache service is running on the webservers
- TestCase for checking curl is working for webservers
RUN python -m pytest -v --ansible-inventory=inventory/hosts --connection=ansible tests/webservers_test.py --force-ansible
Result
For testing Loadbalancer
I have written 5 testcases for testing loadbalancer
- TestCase for having root prevelleges
- TestCase for validating haproxy is installed on the loadbalancer
- TestCase for validating the haproxy service is running on the loadbalancer
- Testcase for validating loadbalancer is working fine even one node is down
- TestCase for checking curl is working for loadbalancer
RUN python -m pytest -v --ansible-inventory=inventory/hosts --connection=ansible tests/loadbalancer_test.py --force-ansible
Result
- Brief summary of what you liked about your solution
This solution demonstrate knowlegde on lot of DevOps tools like Vagrant, Terraform, Ansible, HAproxy and Testinfra(python).
- Brief summary of what you disliked about your solution
Instead of assigning fixed ips to the VMs, I could have used dynamic ips. But I haven't hardcoded ips anywhere. Ansible playbook read ips from inventory file only.
- Configurable Round Robin / Sticky Load Balancer
Yes, I have added both the configurations for loadbalancer. Based on the value of sticky_session passed by the user, playbook will set haproxy.cnf file in loadbalance vm. Basically I have two jinja2 templates templates/haproxy.cfg.j2 for roundrobbin and templates/haproxy_sticky_session.cfg.j2 for stickysession.
- Return instance identifier of your webserver in addition to “Hello World”
Yes, I have included webserver identifier and webserver ip along with Hello World. My webpages will return
Hello World from <webserver1/webserver2>, ip: <webserver ip>