signalfx/splunk-otel-collector-chart splunk-otel-collector-0.138.1

on GitHub

This Splunk OpenTelemetry Collector for Kubernetes release adopts the Splunk OpenTelemetry Collector v0.138.0.

💡 Enhancements 💡

• agent: Update Istio autodetect rules. The detection logic has been expanded to identify Istio components more reliably. (#2132)
  This change may impact billing, as the detection rules have been updated to be less restrictive.
  Metrics will now be scraped from istiod, Istio gateways, and Istio sidecars (istio-proxy containers.)
  Metrics were previously only being scraped from istiod.
• agent: Filter incoming metrics from Prometheus receiver when Istio autodetection is enabled. (#2141)
  The prometheus receiver is now being used when autodetect.istio: true instead of prometheus_simple.
  The reason for this change is that the prometheus receiver supports filtering metrics being scraped by
  the receiver. Filtering has been introduced to only scrape metrics that are included in default content.
• clusterReceiver: Disabling attributes not needed for k8s cluster receiver metrics (#2157)
  The following attributes have been disabled to reduce cardinality:
  • k8s.container.status.last_terminated_reason
  • k8s.hpa.scaletargetref.apiversion
• clusterReceiver: Enable k8sobjects for pods, in pull mode at 6h frequency (#2147)
• clusterReceiver: Added k8s.pod.qos_class (#2144)
• clusterReceiver: Enable metrics (k8s.container.status.reason, k8s.pod.status_reason, k8s.node.condition) for cluster receiver (#2151)
  Enable the following metrics by default in K8s cluster receiver configuration:
  • k8s.container.status.reason: this metric gets containers in CrashLoopBackOff, ImagePullBackOff state, for example, using k8s.container.status.reason attribute.
  • k8s.node.condition: this metrics gets Nodes in MemoryPressure, for example, with condition attribute
  • k8s.pod.status_reason: this metric helps detect Pods in status Evicted, for example
• operator: Add support for instrumentation.spec.defaults field in Helm values to configure default instrumentation behavior. (#2152)
  Users can now configure the defaults.useLabelsForResourceAttributes option via Helm values, allowing
  Kubernetes labels (e.g., app.kubernetes.io/name, app.kubernetes.io/version) to be used as resource attributes
  in auto-instrumented applications. Read more here.
  Example usage:

  instrumentation:
    spec:
      defaults:
        useLabelsForResourceAttributes: true

🧰 Bug fixes 🧰

• agent: Fix kubeletstats receiver failing on GKE Autopilot clusters running Kubernetes 1.32+ (#2159)
  • GKE Autopilot clusters running k8s 1.32+ disable the insecure kubelet read-only port (10255) by default.
  • The receiver in GKE Autopilot now uses the secure kubelet endpoint (port 10250) with serviceAccount authentication.
  • Add RBAC permissions to ClusterRole for restricted nodes/pods endpoint in GKE Autopilot