Skip to main content

1 · Telemetry sources

infra-agnostic

Kubernetes workloadspods · kubelet · cAdvisor · eventsscrape /metrics · OTLPVMs / bare metalnode_exporter · journald · app procsOTLP gRPC :4317 / HTTP :4318Cloud servicesmanaged DBs · LBs · cloud logs APIsfile tail · journald · k8s apiGrafana AlloyDaemonSet on k8s · systemd unit on VMs

Sources​

Kubernetes workloads​

Pods, kubelets and the control plane emit all three signals. Alloy runs as a DaemonSet and discovers targets via the Kubernetes API.

  • Metrics: /metrics endpoints, kubelet, cAdvisor, kube-state-metrics
  • Traces: apps instrumented with OTel SDKs export OTLP to the node-local Alloy
  • Logs: container stdout/stderr files under /var/log/pods
protocolOTLP :4317/:4318 · HTTP /metrics · file tail

VMs / bare metal​

Non-Kubernetes hosts run Alloy as a systemd unit; same pipelines, static or file-based service discovery.

  • node_exporter for host metrics
  • journald + flat files for logs
  • OTLP from local app processes
protocolOTLP · HTTP /metrics · journald

Cloud services​

Managed services that can't run an agent are pulled or pushed through cloud-native integrations.

  • CloudWatch / Cloud Monitoring exporters → metrics
  • Cloud log sinks (e.g. Kinesis/PubSub) → Loki via Alloy
  • LB / gateway access logs
protocolcloud APIs → Alloy pipelines