5. Developing viral-ngs¶
5.1. Testing easy_deploy with Vagrant and Ansible¶
By default, easy_deploy installs a version of viral-ngs by cloning the github repository, which makes it difficult to test local changes to the playbook. Testing the easy_deploy setup locally can be done using Vagrant and the Ansible playbook with some custom commands. To do so, we need to take the following steps:
$ cd easy_deploy
- Run the easy deploy script to bootstrap the VM and answer the prompts
- From now on, run Ansible playbook manually for provisioning: http://docs.ansible.com/ansible/guide_vagrant.html#running-ansible-manually.
- Add deploy=sync or deploy=archive to the –extra-vars of the Ansible playbook command
$ ansible-playbook ... --extra-vars=deploy=sync
To test playbook changes in a local repository, the choice of deploy=sync or deploy=archive changes the strategy used to “deploy” viral-ngs into the Vagrant VM.
The deploy=sync option creates a symlink to the synced folder containing the root of your viral-ngs git repo. Therefore, any tool installation or other changes will be reflected on both the host and the guest machine. This is desirable for fast iteration of changes, but makes it difficult to isolate the host’s viral-ngs installation from the installation on the guest VM.
The deploy=archive option performs a git archive on the host’s viral-ngs repository and untars it into the project directory. This is a clean install each time, which can be time consuming due to the need to reinstall all dependencies, and only the current HEAD commit will be reflected in the guest. No uncommitted/dirty changes will be picked up using this method. This is more ideally suited for a finalized clean test of the playbook.