Dashboard
The Dashboard (/dashboard) is the first page a platform administrator sees after signing in. It answers one question quickly: is my cluster healthy, and if not, where should I look? Tenant users land on their own home page (/home), covered in Your Dashboard.
The page has four sections, all refreshed together every 60 seconds.
Cluster overview
Six cards summarize the cluster:
| Card | What it tells you |
|---|---|
| Health Score | An overall score out of 100 and the number of active issues. See how it's calculated. |
| Ready Nodes | Ready nodes out of total nodes. |
| Restarting Pods | Pods that are restarting, alongside running and total pods. |
| CPU Usage | Average CPU utilization across nodes. |
| Memory Usage | Current memory utilization. |
| Monthly Cost | Projected monthly cost of the cluster's nodes, based on their current hourly rates. |
The header shows when the data was last refreshed.
Active health issues
When something needs attention, it appears here as a card with:
- a severity badge (critical, warning or info);
- the resource affected;
- a message explaining the problem;
- a suggested fix, when KubeOpera has one.
The most important issues are shown first. Select View all to see every open issue.
Issues often have a matching AI recommendation or an open incident. Follow the links on each card, or ask the AI Chat "what should I do about this?".
Resource utilization
CPU and memory charts show how usage has changed over time, so you can spot trends — a slow memory climb, a daily CPU peak — before they become incidents.
Observability
Four panels give you a quick read on your workloads:
| Panel | Shows |
|---|---|
| Latency | Request latency across instrumented services. |
| Pod Status | Running, pending and failed pods. |
| Network Traffic | Inbound and outbound traffic. |
| Error Count | 5xx errors across services. |
Latency, network and error panels come from Prometheus. For them to show data, your workloads should expose Prometheus metrics — see Monitoring for how to instrument an application. When a panel is empty, it tells you why: whether there is simply nothing to report (for example, no errors), or whether the workload isn't instrumented yet.
What to do next
- A cluster looks unhealthy? Open Clusters for per-node detail, or Incidents to see whether one is already open.
- Something unusual in the charts? Analytics shows detected anomalies and forecasts.
- Want an explanation? Launch an AI investigation.