Red Hat OpenShift 5 adds AI-driven observability with autonomous troubleshooting, Perses dashboards

Red Hat OpenShift 5 ships with a major observability overhaul centered on Red Hat OpenShift Lightspeed, an AI-powered layer that automates cluster troubleshooting and dashboard creation. The release introduces three capabilities: autonomous troubleshooting agents that investigate alerts and stage remediations pending human approval; an intelligent assistant that synthesizes metrics, logs, traces, and events into structured root-cause summaries; and natural-language-generated Perses dashboards built without PromQL knowledge. These features require specific operator versions—Lightspeed operator 2 for autonomous agents, Lightspeed operator 1.1.5 or later for the intelligent assistant, and cluster observability operator 1.6 for AI-generated dashboards.

The cluster observability operator 1.6 also brings operational improvements including default Thanos data source wiring, console-based Perses project creation, and bundled plug-ins for air-gapped deployments. On the monitoring infrastructure side, OpenShift 5 changes defaults for Prometheus remote-write versioning and enables the node-exporter NVMe subsystem collector by default, which exposes NVMe over Fabrics path health metrics required for storage degradation alerts. A technology preview ClusterMonitoring custom resource begins migrating configuration from the unvalidated cluster-monitoring-config ConfigMap.

What changes in OpenShift 5 observability stack

• Autonomous troubleshooting agent: Configured via Alertmanager adapter, automatically investigates firing alerts by inspecting cluster state, drafts diagnoses, and stages remediations in a decision queue; all state changes require explicit admin approval within sandboxed pods under dynamic RBAC, with automatic human escalation on failure.
• Intelligent assistant mode: Synthesizes metrics, events, pod logs, and distributed traces into structured summaries for manual troubleshooting via console side panel or CLI, available in production with Lightspeed operator 1.1.5+ on OpenShift 4.22+.
• AI-powered Perses dashboards: Natural-language queries generate on-demand charts through cluster observability operator 1.6 and Red Hat's Perses build, with AI-Assisted badges enabling audit of underlying PromQL, time ranges, and metrics before pinning to production dashboards.
• Cluster observability operator 1.6: Adds out-of-box Thanos default data source, UI-based Perses project creation without YAML manifests, and air-gapped support via bundled container image plug-ins.
• Monitoring infrastructure defaults: Prometheus remote-write defaults to V1.0 message version per endpoint with optional V2.0 metadata support; node-exporter enables NVMe subsystem collector by default exposing controller visibility and path states for CSIVolumeNVMePathDegraded and CSIVolumeNVMePathLost alerts.

Why it matters

For Kubernetes platform teams and SREs running OpenShift, these changes reduce mean time to resolution by automating alert triage while preserving enterprise audit requirements through mandatory approval gates. The shift toward natural-language interfaces for PromQL and dashboard generation lowers the observability skill barrier, though organizations in regulated or disconnected environments will particularly benefit from the air-gapped Perses bundling and explicit AI-generated content labeling for compliance review.