Deploying Django on AWS with Zero Downtime

Deploying Django on AWS with Zero Downtime, a comprehensive guide available on the Djangix blog, demonstrates an efficient method to maintain uninterrupted service while updating Django applications. The strategy involves utilizing a blue-green deployment approach on a single server, eliminating the need for Kubernetes or an additional machine. This involves maintaining two complete copies of the Django app, running one at a time, and directing traffic through Nginx.

Each colour of the deployment process has its distinct code directory, virtual environment, Gunicorn service, and local port. However, they all share the same database, media storage, and environment configuration. To deploy, the inactive copy is updated, checked, and migrated, and then started and tested for health before any user sees it. Once the new colour responds correctly, Nginx is pointed to its port, and the proxy is reloaded.

The beauty of this method lies in its rollback capability. If the new release misbehaves, the proxy can be switched back to the previous colour, and the reload initiated within seconds, without the need for a rebuild or frantic restore. This approach not only simplifies the deployment process but also ensures faster rollback times, making it a highly reliable practice.

Key considerations include keeping secrets out of either app directory, preventing collectstatic from overwriting files served by the live copy, and treating database changes that the old code cannot handle as a separate issue, as an instant code rollback cannot undo incompatible migrations.