Dashboards Reference
The dashboard has nine dedicated monitoring pages under Monitoring in the sidebar, each backed by its own real API route rather than being generic widgets over one shared data source. This page is a map of what each one covers, since it's easy to guess wrong from the label alone.
Performance and APM are the closest to each other and the ones most likely to be confused: Performance shows infrastructure-level metrics (CPU, memory, network, pod status) sourced from whichever backend METRICS_SOURCE selects, while APM shows application-level metrics (request latency, throughput, error rate per service) sourced specifically from apm-gateway, regardless of METRICS_SOURCE. If you're trying to answer "is my node under memory pressure," that's Performance; if you're trying to answer "is this specific service slow," that's APM.
Cost breaks down spend by namespace and workload. Security surfaces posture findings and vulnerability counts from security-api. SLOs tracks compliance and error-budget burn per service against configured objectives. Traces renders distributed traces end to end, backed by tracing-gateway. Actions shows the automated remediation log — what the agentic AI layer has actually done, not just recommended.
Optimizer and AI Optimizer are also easy to conflate: Optimizer is the tenant-facing resource-right-sizing view (over- and under-provisioned workloads, with suggested changes); AI Optimizer is a narrower page specifically surfacing kubeopera-ai's own health-score and optimization output.
Every one of these pages polls its backend on an interval rather than pushing updates over a websocket — refreshing your browser (or just waiting for the next poll) is always safe if a number looks stale, there's no manual "refresh cache" action needed anywhere.