Skip to main content

Case Study: How Wenrix Made Schema Migrations Reliable with Atlas

· 4 min read
Noa Rogoszinski
Noa Rogoszinski
DevRel Engineer

"Atlas just works behind the scenes. We don't need to pay attention to it – we can trust that it's getting the job done."

– Armon Avrahamy, CTO and Co-Founder, Wenrix

Company Background​

Wenrix is the AI infrastructure for profitable growth in air. Since 2018, leading travel agencies have relied on Wenrix to help them grow revenue, protect margin, and scale more efficiently in an increasingly complex and competitive air retail landscape.

Keeping Up With a Growing Stack​

As Wenrix grew in its beginning stages, so did the surface area for schema changes. The team runs multiple PostgreSQL databases segmented by function and product, deploys on Kubernetes, and ships multiple deployments per day. Keeping application code and the database state aligned at this pace became a difficult operational problem, and the migration workflow in place was not built for it. As Wenrix expanded its platform and accelerated release velocity, database schema management became an increasingly important part of maintaining engineering efficiency.

In the early years, migrations were entirely manual. While the team eventually integrated tools like go-migrate with custom CI scripts, the operational complexity of managing schema changes across multiple services continued to grow. “We deployed two to three times a day and needed a more scalable way to manage schema changes across environments,” said Armon Avrahamy, CTO and co-founder. Coordinating manual steps such as timing, verification, and validation of schema safety became a significant bottleneck as the team scaled.

Manual coordination became increasingly difficult to sustain at the company's deployment velocity. Managing data at this scale required moving past manual processes toward an automated approach. Ultimately, Wenrix needed a migration workflow that could keep pace with how the company actually operated.

Searching for a Solution​

Armon began evaluating migration tools that could impose more structure without compromising on efficiency. Python-specific options like SQLAlchemy were an early consideration, but not every service in Wenrix's stack used the same ORM, and tying schema management to a single application framework would have meant different migration paths for different parts of the system. That inconsistency was exactly what they were trying to avoid.

They also looked at established SQL migration tools such as Flyway, but found the fit wrong for a different reason. "Other solutions seemed too complicated for our specific use case," Armon said. Wenrix did not need a heavyweight, Java-centric workflow layered on top of a Python and Kubernetes environment. They needed something that could sit alongside their existing deployment pipeline and treat schema changes the same way everywhere.

Choosing Atlas​

In 2024, Wenrix turned to Atlas. Its architecture matched how Armon wanted to think about schema changes. "I was drawn to the similar concept to Terraform," Armon said, describing a declarative model where the desired state is defined in code and the tool plans what needs to happen to get there.

With guidance from the Atlas team, Wenrix moved from a purely declarative workflow to versioned migration files generated with a pre-commit script, paired with the Atlas Kubernetes Operator to apply changes as part of their deployment pipeline. That combination gave them explicit, reviewable SQL migrations and automated enforcement at deploy time.

The Outcome​

Choosing Atlas gave Wenrix a migration workflow built for the pace at which they actually ship. Armon pointed to several benefits:

  • Deployment Gating. If a migration is unsafe, the system automatically prevents the deployment. This ensures that only validated schema changes reach production, maintaining environment consistency without manual intervention.
  • Minimal Overhead. Beyond reliability, the workflow faded into the background. "Atlas just works behind the scenes," Armon said, requiring very little ongoing attention from developers.
  • Low-Friction Schema Changes. Versioned migration files generated at commit time mean engineers change one file and the rest of the pipeline handles planning, review, and apply. "Running the migration is much easier to do now. You just edit one file, start the pipeline, and it's done," he added.
  • Healthier Data Modeling. When schema changes require significant effort, teams naturally become more conservative about structural changes. Now that migrations are seamless, teams can prioritize clean, structured data models that evolve alongside the application code.
  • Operational Scalability. Schema changes run automatically through the Kubernetes operator as part of every deployment, removing the operational complexity of managing manual coordination across multiple services.

Getting Started​

By managing your schema as code and integrating Atlas into your CI/CD or Kubernetes deployment pipeline, you can automate migration generation and block unsafe changes before they reach production.

If your team ships frequently and relies on manual migration steps, or if managing the alignment between application code and database schema has become an operational bottleneck, Atlas can help streamline your delivery pipeline.