Kubernetes onboarding

Diting supports Kubernetes 1.26 through 1.33. Clusters, nodes, workloads, network, storage, events and logs share one navigation.

Access

• kubeconfig: credentials stored by the platform, authorized per cluster
• In-cluster agent: deployed inside the cluster, for restricted networks

Both are read-only against the Kubernetes API. Workloads are not modified.

Resource coverage

• Clusters and nodes
• Namespaces and projects
• Deployments, StatefulSets, DaemonSets, Jobs
• Services, Ingress, HPA, NetworkPolicy
• PV, PVC, StorageClass
• ConfigMaps, Secrets
• Events, logs, metrics

Permissions

Grant by cluster, namespace and host group. Users only see resources and data inside their grant.

Relationship to hosts

Kubernetes resources share one graph with hosts. Which node a pod runs on, which database it calls and which service it affects are all connected.

Chaos

Three CRDs are provided:

• chaosexperiments
• chaosexecutions
• chaospolicies

Experiments run inside the cluster and report results back to the platform.

Verify

1. The cluster appears with an available state
2. Node and workload counts match kubectl
3. Events and logs filter by workload
4. A pod links to its host and trace

Limits

• Fully managed clusters with no API access cannot be onboarded
• Clusters that forbid an agent provide metrics only

Related: Traces and topology, Access, audit and tenancy.