System commands
Twenty-one of the 43 commands.
| If you want | Use |
|---|---|
| One answer on whether the node is well | health |
| Connectivity and DNS probes run by the node itself | diagnosis |
| Kubernetes workload containers | crictl |
| The platform containers: vpm, argo, site-console | Docker |
| Service logs, including registration | journalctl |
The node runs two container runtimes
Section titled “The node runs two container runtimes”This surprises people, and it changes which command answers your question.
- Docker runs the platform containers —
vpm,argo_watch,site-console. - CRI-O runs the Kubernetes workloads, reached through
crictl.
Both are live simultaneously, both return real output, and neither substitutes for
the other. docker-ps showing three containers is not a partial view of a broken
node; it is the complete list of what Docker runs. If you are looking for vpm, it
is a Docker container. If you are looking for a pod, it is CRI-O.
Two commands use a different transport
Section titled “Two commands use a different transport”health and diagnosis are GLOBAL scope: they are GET requests to a global
path, return JSON rather than terminal text, and are not reachable through
exec-user — sending them there answers command not supported, which reads as
though the command does not exist. See the debug API.