Health checks should match the app, not just the port
A good health check tells the load balancer whether the backend can actually serve the workload, not merely whether the server is powered on.Practical guidance
- Use an HTTP endpoint when the application can expose one safely.
- Use a TCP check only when a deeper application check is not possible.
- Keep the check fast enough to catch failure but stable enough to avoid flapping.
- Review timeout and interval settings during traffic tests.
Common mistakes
- Probing an endpoint that is too shallow to reflect app health
- Using check intervals that are too aggressive for the workload
- Forgetting to update the health check after the application path changes