
Data Integration Platform CI/CD (Azure Data Factory)
Reworked CI/CD for an internal Azure Data Factory platform, replacing manual promotions with a Dev → Preprod → Prod pipeline, clear environment-specific configuration, and approval gates for safer releases.
DevOps Engineer2025 · ~3 monthsInternalAzure Data FactoryCI/CDDevOpsData Engineering
Data Factory deployments between Development, Preproduction, and Production were mostly manual and inconsistent, which slowed releases and caused avoidable issues.
- Data Factory wasn't following a clean CI/CD approach.
- Manual promotion caused drift and repeat deployment errors.
- Environment-specific values were effectively hard-coded, making changes risky.
- Database and data engineers were getting blocked by deployment problems.
- Introduce a repeatable CI/CD process aligned with Microsoft's recommended approach for ADF.
- Promote changes safely across Dev → Preprod → Prod.
- Parameterise environment-specific values without duplicating code.
- Reduce deployment issues for engineers working on pipelines and data pipelines.
- Set up a reliable promotion path for ADF across environments.
- Removed repeated deployment blockers for data and database engineers.

- Azure Data Factory instance per environment.
- Managed Identity used to connect ADF to dependent Azure resources.
- Linked services parameterised to support different environments safely.
- ARM templates generated from ADF for controlled deployments.

- CI pipeline generates ARM templates from the Dev Data Factory (adf_publish flow).
- Environment-specific parameters applied during deployment (Preprod/Prod).
- Automated deployment to Preprod to keep environments in sync.
- Approval gates for Preprod (test team) and Production (DevOps team).
ARM-based CI/CD (supported ADF approach)
ARM templates require additional configuration, but the deployment model is predictable and well-supported.
Auto deploy to Preprod, gate Production
Production changes take an extra step, but releases are safer and easier to track.
- Managed identities used for Data Factory access to Azure resources.
- No secrets stored in source control or pipeline definitions.
- Secrets pulled from Azure Key Vault via linked services instead of being stored in pipelines or config.
- Production deployments restricted via approvals and permissions.
- Deployments followed the same process across environments, reducing environment-specific failures.
- Standard promotion flow reduced deployment-related incidents.
- Rollback is straightforward by redeploying the last known-good template.
- Reliable CI/CD pipeline for Azure Data Factory deployments.
- Fewer deployment errors and less configuration drift across environments.
- Improved productivity for database and data engineers by removing deployment blockers.
- Add automated validation checks for ADF changes before deployment.
- Improve pipeline output with clearer summaries (which pipelines/linked services changed).
- Expand monitoring and alerting around ADF execution failures.