Built-in health checks should have configurable failure statuses
j-mok opened this issue · 3 comments
Problem:
ModulesHealthChecker is hard-coded to report failure as Degraded (here and here). While this is ok in some scenarios, it is not in others, including ours. We deploy the platform with a rigid set of module versions and expect all of them to load successfully before we can deem a deployment successful. A failure to load one module warrants deployment abort. In our deployment pipeline we poke the /health endpoint to see if it return 200 before swapping with production. But since ModulesHealthChecker reports Degraded in case of a module load error, the endpoint maps that to 200, not 503, and we're forced to parse the health report to see the status of the modules.
Solution:
Make failure statuses independently configurable for built-in health checks (ModulesHealthChecker and CacheHealthChecker).
@j-mok I agree with you, health should return 503 in case of a module load error, We will fix it in coming release.