SYS.DOCS // DOCS

Kubernetes Clusters Overview

The Overview tab is the summary page for a cluster. It shows whether the cluster needs attention, how its nodes and workloads connect, how much capacity is in use, and how the cluster is configured.

Summary

The strip at the top of the page shows:

Location: the region the cluster runs in.
High Availability: whether the control plane runs on three servers instead of one.
API Endpoint: the Kubernetes API server address, with a copy button. Clusters with backup endpoints list them in a dropdown next to the primary one.
Cost: the monthly run rate and the amount spent this month.
Access: download your personal Kubeconfig and the cluster’s SSH Key. A green dot marks a credential you have already downloaded. See Cluster Access & Security for how credentials are issued and revoked.

Health

The Health line shows the cluster’s status: Healthy, a count of warnings or issues, Unreachable or Disconnected, or Unknown while Edka has no node data yet. It also shows when the in-cluster edka-agent last reported and its version. Diagnostics opens the full Cluster Diagnostics page.

When Edka cannot reach the cluster, the banner at the top of every cluster page gives the cause, when the cluster was last reachable, and Check now.

When the cluster has warnings, the Health panel sits below the topology graph and lists up to four of them. Each warning opens the section or page with its detail:

Warning	Raised when	Opens
Pods	Pods are crashing or failed, or have been starting for more than 3 minutes	Diagnostics, Pods Requiring Attention
Nodes	Nodes are not ready, or cordoned and not accepting new pods	Explorer → Nodes, filtered to nodes with issues
etcd	An etcd member reports no leader	Diagnostics, K3s Signals
Certificates	K3s certificates expire within 30 days, or every recent warning event belongs to a cert-manager resource	Diagnostics, K3s Signals, or Domains for cert-manager events
Snapshots	etcd snapshots failed on the control plane	Diagnostics, K3s Signals
Events	Kubernetes reported warning events, grouped by reason, for example `BackOff ×3 (Pod)`	Diagnostics, Kubernetes Events
Backups	A scheduled etcd backup is missing from the backup catalog	Cluster Settings → Backups
Agent	`edka-agent` stopped reporting, failed to install, or reports degraded	Cluster Settings → Add-ons

A Monitoring note means Edka’s own connectivity check is running late, so outage detection can be delayed. It is listed but does not change the cluster’s status.

Infrastructure topology

The topology graph shows gateways on the left and node pools beside them. Pools sit side by side in rows that fit the panel width, each pool a frame holding its node cards, and each node card lists the workloads running on that node.

Lines run from each gateway to the pools it sends traffic to. A line turns green when a gateway route points at a workload in that pool, and moves while the gateway has requests or open connections.
Node cards show CPU and memory usage against what the node offers to pods, and how many pods it holds against its pod limit. A tick on the CPU and memory bars marks requested resources, which is what the scheduler uses to place pods. Hover a bar for the exact numbers.
Each workload row shows one square per pod on that node. Green pods are running, amber pods are still starting, red pods are crashing, failed, or have been starting for more than 3 minutes, and grey pods have completed. Workloads with problems are listed on the card.
Healthy workloads fold into one +N healthy workloads row per node. Click Show to list them and Hide to fold them again. A gateway icon marks workloads a gateway route points at.
Gateway cards show their endpoint and traffic from Envoy Gateway: requests per second, open connections, 5xx rate, and latency.
Click a gateway, pool, or node card, or a workload row, to open its details. A workload’s details list its pods, failing pods first, with buttons for their logs and exec.
Filter by namespace, or search for a node, workload, or pod name. The issue count in the panel header switches the graph to only the workloads with problems.
The plus and minus buttons in the graph controls zoom in and out, and the fit button fits the graph to the panel width.
The chevron next to the refresh button collapses the panel to its header, which keeps the counts. Your browser remembers the choice.

While the cluster is unreachable, the graph is hidden until it reconnects.

Capacity

Capacity shows one row each for CPU and memory, summed across all nodes:

Used: live usage. It needs the metrics-server add-on.
Requested: the sum of container resource requests.
Limits: the sum of container resource limits. A value above 100% means limits add up to more than the nodes can supply. Hover it to see the multiple.
Allocatable: what the nodes offer to pods.

The bar fills with usage and marks requests and limits. The header shows how many nodes are ready and how many pods are running and completed. The numbers refresh every minute while the page is open, and Refresh updates them immediately. By workload opens Resources, which breaks usage and requests down per workload. While the cluster is unreachable, the header shows the time of the last update.

Configuration

The configuration panel lists the cluster’s settings in three columns. The Kubernetes and Networking headings open the matching page in Cluster Settings.

Kubernetes

Version, with the next upgrade next to it when one is available: the newest patch of the current minor version, then the next minor version. See Update a Cluster.
OS, Control plane size, CNI, CNI encryption, and Secrets encryption.
etcd backups: the backup schedule, in UTC.
Last backup: when the newest etcd backup finished. It turns amber when a scheduled backup is missing. See Cluster Backups and Restore.

Networking

Public IPv4, Public IPv6, and the Private network subnet.
NAT gateway size and availability.
Host firewall.
SSH allowed and API allowed: the source networks allowed to reach SSH and the Kubernetes API.

Integrations

GitOps: the connected repository, or Not configured.
Registry: the connected registry, or Not configured.
Deployments: the six most recently active deployments with their status. Each one opens the deployment.

Satellite clusters show the server’s IP address under Networking and skip the rows that only apply to clusters Edka provisions on Hetzner.

Add-ons

Add-ons lists the add-ons on the cluster with their versions. Add-ons that are installing, upgrading, or failed show their state. An arrow marks an add-on with a newer version in the catalog, for example v1.21.1 → v1.22.0, and the heading counts the available updates. The heading opens Cluster Settings → Add-ons, where you apply updates.

Operational workspaces

The Overview is the cluster summary. The dedicated workspaces cover day-2 operations:

Cluster Observability: metrics, retained logs, diagnostics, and VictoriaMetrics or VictoriaLogs management
Databases overview: PostgreSQL, MySQL, Valkey, and ClickHouse inventory, restore, runtime logs, insights, and engine-specific settings
Cluster Backups and Restore: etcd backup catalog, encryption, rollback, deleted-cluster recovery, and restore-as-new workflows
Cluster Explorer: live Kubernetes resources, Helm releases, manifests, logs, nodes, and pod debug sessions
GitHub Actions Runners: private ARC runner scale sets, runtime pods, and CI logs for this cluster