This case study will walk through running a SaaS product at scale on Kubernetes.

Running at scale on Kubernetes sounds simple—until you try to do it with stateful workloads, strict uptime requirements, and thousands of dedicated customer environments. Since 2019, Octopus Deploy has operated its SaaS offering entirely on Kubernetes, now spanning 1,700 customer instances across 11 clusters in three regions.

Each customer gets a dedicated application, database, and file share, all sharing the same codebase as our self-hosted product. What started as a brownfield migration quickly became a hard lesson in balancing reliability, cost, and velocity.

This session will share the lessons from six years of running and evolving a stateful SaaS platform on Kubernetes.

You’ll hear about:
- Why we rebuild clusters for new Kubernetes versions instead of upgrading in place
- The ongoing trade-offs between over-provisioning and cost control
- How managed services like Azure SQL, Key Vault, and File Storage simplified operations—but not without challenges
- The journey to eliminate maintenance windows and rewrite a critical execution pipeline, all while maintaining 99.99% uptime

If you’re planning to run a SaaS product on Kubernetes—or are already doing so—this talk will give you practical insights and cautionary tales from the trenches.

Target Audience: Developers, SREs, and DevOps engineers running or planning to run SaaS products on Kubernetes, especially those managing stateful workloads or multi-tenant environments.

Takeaways:
- Practical lessons for running stateful SaaS on Kubernetes at scale
- Strategies for upgrades, maintenance windows, and resource management
- How managed cloud services can simplify operations
- Techniques to maintain high SLOs and deliver features safely

Session Format: Talk with slides and real-world examples from six years of operating 1,700 SaaS instances across 11 clusters.