nodefony-devops

Start from what the Nodefony generator already renders: a Dockerfile template, Compose file, CI workflows, and Kubernetes manifests including a migrate job. Inspect the image's built-in capabilities first, such as /livez and /readyz probes and a 15-second shutdown drain, before writing any custom YAML.

The base preset renders Dockerfile.tpl plus GitHub and GitLab CI templates; the complete preset adds compose.yaml.tpl, a production workflow, and a migrate-job.yaml.tpl. The it.db template variable decides whether a server database exists, which controls replica count.

Yes, use `podman kube play deploy/` to run manifests without a cluster, or create a disposable kind cluster with `kind create cluster` and apply the Restricted pod-security label. Note that Podman does not replay admission policies, Ingress, or rollouts.

A sqlite database forces a single replica and a Recreate strategy. A RollingUpdate on a ReadWriteOnce volume waits for a disk the old pod still holds, leaving the deployment Pending forever without an error naming the cause.

Behind a proxy terminating TLS, the app receives plain HTTP with an X-Forwarded-Proto header. Without trustProxy set, the detected scheme is http and __Host- cookies silently fail; set trustProxy in the manifest.

Exit code 137 means the container was killed, so graceful shutdown did not occur; the only acceptable verdict is 0. For example, `podman stop` defaults to 10 seconds, below the framework's 15-second drain, so in-flight requests are lost unless the timeout is raised.