Kubernetes¶
The Kubernetes console lets you browse and operate workloads in your account’s namespace on the platform cluster. You do not paste a kubeconfig for normal use — the console talks to the cluster on your behalf.
| Item | Value |
|---|---|
| Console | Kubernetes → /console/kubernetes |
| Scope | Your account namespace (acc-{shortId} style). Platform admins also see system namespaces, never other tenants’ acc-* namespaces |
What you can do¶
- List namespaces you are allowed to see
- Browse Deployments and Pods
- Inspect status, restarts, and events
- Adjust scaling / HPA where exposed
- View and edit configuration
- Apply YAML for advanced custom resources (validate before apply)
Console walkthrough¶
Find your workloads¶
- Open Kubernetes.
- Select your account namespace.
- Open Workloads / Deployments to see replica counts and readiness.
Pods¶
Open a deployment → Pods:
- See running pods, restarts, node placement
- Use this when an app is “up” in Applications but traffic fails — verify pods are Ready
Scaling¶
Open Scaling:
- Set desired replicas
- Configure or inspect Horizontal Pod Autoscaler when available
- Use health / metrics / timeline panels when your console version shows them (in-console Info explains each chart)
Configuration¶
View environment, mounts, and related settings. Prefer changing Applications Settings when the workload was created from an Application template, so you do not drift from the app model.
Apply YAML¶
For power users:
- Paste a manifest.
- Validate.
- Apply only when validation succeeds.
Avoid applying changes that fight an operator or Application controller.
Relation to other services¶
| Service | Relationship |
|---|---|
| Applications | Higher-level deploy; uses Kubernetes under the hood |
| MDB / Redis | May show related pods for debugging; manage day-2 via their own consoles |
| Functions | Separate runtime; not edited as raw Deployments for normal use |
Tips and pitfalls¶
- Account isolation is namespace-based — you will not see other customers’ workloads.
- Cluster administration (nodes, Traefik, storage classes) is a platform operator concern.
- If Apply fails, read Events on the resource before retrying.
- Scaling to zero may be limited by platform policy for some workloads.