I work on the deployment pipeline for TiBillet Gala instances, an open-source ticketing, point-of-sale and cashless payment system. The project connects AWS infrastructure, PostgreSQL databases, Docker images and controlled production releases.
Deploy connected services while isolating each event's data and keeping the previous installation until validation is complete.
A Gala instance combines Lespass, Fedow and Laboutik: ticketing, federated payments and point of sale. The proposed architecture isolates each gala on its own EC2 instance, with dedicated secrets and backups. Terraform describes AWS resources and scripts control launch operations.
Terraform defines instances, access, storage, image registries and pipelines, using remote state and scoped IAM permissions.
The Test pipeline builds and publishes immutable images to ECR. A validated image is a release candidate, not yet a production deployment.
A release manifest pins service digests. Production is gated by human review, runtime secrets and preflight checks.
Scripts cover PostgreSQL backup, startup, health checks and rollback. Migration remains gradual until the entire flow is validated.
The pipeline contains real resources and scripts for traceable releases. Releases have been pinned and tested on Smoke; the complete target flow and cutover are still being developed.
This work forces me to think about deployment, data, payment and rollback together so the team can operate the system without disrupting the existing service.