grafana/k8s-monitoring-helm k8s-monitoring-4.2.0

on GitHub

What's changed

• Add hostMetrics.linuxHosts.source (node-exporter (default) or alloy). With alloy, Linux host metrics are collected directly by the assigned collector via prometheus.exporter.unix, requiring no Node Exporter deployment. This needs a privileged DaemonSet that mounts the host filesystem, configured with the new presets: [linux-host-monitor, daemonset]. The nodeLabels enrichment is not yet supported with source: alloy. (#2660) (@petewall)
• Introduce Data Processors, a method for transforming data between features and destinations. Initial processors include the custom processor.
• Add the kubernetesEnrichment data processor, which copies Kubernetes namespace and pod labels and annotations onto metrics, logs, traces, and profiles across all supported ecosystems. (@petewall)
• Add the ec2Enrichment data processor, which uses discovery.ec2 to copy AWS EC2 instance tags onto Prometheus metrics, Loki logs, and Pyroscope profiles, matching telemetry to its instance by the node label. (#2678) (@petewall)
• Deprecate the metricEnrichment option on the Prometheus destination in favor of the kubernetesEnrichment data processor. The option remains functional. (@petewall)
• Add a source="spanmetrics" resource attribute to metrics generated by the span metrics connector, making it easier to identify and filter span-derived metrics. (@rlankfo)
• Include Grafana Beyla's Kubernetes cache as deployable from telemetryServices. This is distinct from the Beyla inside the autoInstrumentation feature. (@petewall)
• Fix Prometheus destinations dropping the service.name, service.namespace, and service.instance.id resource attributes from the target_info metric during OTLP to Prometheus conversion. openTelemetryConversion.keepIdentifyingResourceAttributes now defaults to true, so these are retained as the service_name, service_namespace, and service_instance_id labels. Set it to false to restore the previous behavior. (#2722) (@rlankfo)* Include the Grafana SDK Injector, which complements Grafana Beyla's ability to automatically instrument applications. (@petewall)

Chart released at https://github.com/grafana/helm-charts/releases/tag/k8s-monitoring-4.2.0

Full changelog: CHANGELOG.md