Dashboard: Overview · Capacity · Reliability · Cost

Each cluster’s dashboard is organized into sub-tabs built around a scan → dive rhythm: a summary layer up top answers the urgent question at a glance, and the detailed panels below let you drill into anything. Overview and Capacity are always there; Reliability appears when the cluster ships Hubble flow data, and Cost (beta) when it ships OpenCost data. A tab with nothing to show is hidden rather than rendered empty.

Overview — efficiency as the hero

• KPI cards — nodes, pods, active insights and an overall cluster-health ring, with links to the lists behind them.
• Resource efficiency band — a capacity axis per resource with three stripes: used, requested-but-idle (striped), and free. The reclaimable gap is drawn explicitly, with per-resource efficiency scores and right-size callouts where there’s something to reclaim.
• Identity header — breadcrumb plus the cluster’s cloud provider and region, detected automatically from node metadata (they simply don’t render when undeterminable — nothing to configure).
• Workload health at a glance. Completed Job pods are counted apart from Running, so terminal pods don’t inflate the health ring.
• Namespaces as tiles with their workloads and pod readiness, and a recent events feed where each warning has an Ask Kobi affordance (shown when Kobi Copilot is configured).

Capacity — a scan layer over the trends

• Summary strip: peak CPU / memory (with sparklines), a rightsizing opportunity aggregate, and OOMKills in the selected range.
• Trends — CPU, memory, network and filesystem charts overlaid with deploy markers. They read metrics history, so without the agent this block shows a placeholder instead of charts.
• Top workloads by CPU, with ReplicaSet pods collapsed into their Deployment.
• Recent Deploys and Recent OOMKills as scannable tables. OOMKills are enriched with the container’s memory limit and restart count — you can tell a chronic crash loop from a one-off — and each row has an Ask Kobi affordance to investigate.
• Right-sizing rows show per-item reclaimable CPU and memory. With cost data connected, the Cost tab prices those same recommendations.

Reliability — golden signals, honest colors

• Golden-signals strip: error rate (5xx), latency, throughput, and L4 drops, each with a delta against the previous window of the same length.
• Class-pinned colors: 5xx is always red and 4xx always amber, even when one class has no data in range.
• Panels below the strip: error rate split into 4xx and 5xx, top workloads by traffic, top workloads by latency, error hot-spots (sorted by absolute error rate, so a small service that always fails isn’t buried) and network drops (L4 dropped verdicts — the early warning for NetworkPolicy denials).
• Latency is labeled as an average — the agent ships sum + count, not histogram buckets — and a tooltip says so instead of letting you assume a p99.

Shared behavior

The selected time range is shared between Capacity, Reliability and Cost and persists while you drill into resources, but resets to the 15-minute default on a fresh browser session — the daily scan never accidentally fires a 30-day query.

Cost (beta)

Spend, idle cost, cost per pod and money-priced rightsizing, from OpenCost data the agent ships. It is covered on its own page: Cost.