Configure metrics collection

Serverless Observability Stack EDOT Collector

Learn how to configure and customize metrics collection through the Elastic Distribution of OpenTelemetry Collector.

Any application emitting metrics through OpenTelemetry Protocol (OTLP) can forward them to the EDOT Collector using the OTLP receiver. This is the recommended method for collecting application-level telemetry.

The following OTLP receiver configuration turns on both gRPC and HTTP protocols for incoming OTLP traffic:

 # [OTLP Receiver](https://github.com/open-telemetry/opentelemetry-collector/tree/main/receiver/otlpreceiver) receivers: otlp: protocols: grpc: endpoint: 0.0.0.0:4317 http: endpoint: 0.0.0.0:4318 

Configure your application to export metrics using the OTLP protocol, targeting the endpoints provided in the previous example.

The hostmetrics receiver turns on the collection of host-level metrics such as CPU use, memory use, and filesystem stats.

The following configuration collects a standard set of host metrics that aligns with Elastic's Infrastructure dashboards in Kibana:

 hostmetrics: collection_interval: 10s root_path: /proc scrapers: cpu: metrics: system.cpu.utilization: enabled: true system.cpu.logical.count: enabled: true memory: metrics: system.memory.utilization: enabled: true network: {} processes: {} load: {} disk: {} filesystem: exclude_mount_points: mount_points: - /dev/* - /proc/* - /sys/* - /run/k3s/containerd/* - /var/lib/docker/* - /var/lib/kubelet/* - /snap/* match_type: regexp exclude_fs_types: fs_types: - autofs - binfmt_misc - bpf - cgroup2 - configfs - debugfs - devpts - devtmpfs - fusectl - hugetlbfs - iso9660 - mqueue - nsfs - overlay - proc - procfs - pstore - rpc_pipefs - securityfs - selinuxfs - squashfs - sysfs - tracefs match_type: strict 

1. Mounted node's root file system

You must grant access to the /proc filesystem to the receiver by running the Collector with privileged access and mounting /proc and /sys appropriately. Refer to the hostmetrics container use guide (Linux only).

Turning on the process scraper might significantly increase the volume of scraped metrics, potentially impacting performance. Refer to the contrib issue #39423 for discussion.

To ensure compatibility with Kibana's Infrastructure dashboards, include the elasticinframetrics processor in your pipeline:

 service: pipelines: metrics/infra: receivers: - hostmetrics processors: - elasticinframetrics 

Process metrics are turned off by default to avoid generating a large volume of timeseries data. To turn on process metrics, uncomment or add the following section inside the hostmetrics receiver configuration:

 process: mute_process_exe_error: true mute_process_io_error: true mute_process_user_error: true metrics: process.threads: enabled: true process.open_file_descriptors: enabled: true process.memory.utilization: enabled: true process.disk.operations: enabled: true 

You can collect Kubernetes metrics using multiple receivers depending on the type and source of the metrics. Each receiver might require specific Kubernetes permissions and require a deployment as DaemonSets or singletons.

As with host metrics, use the elasticinframetrics processor to ensure metrics align with the Kibana Infrastructure inventory.

The kubeletstats receiver collects resource usage stats directly from the Kubelet's /stats/summary endpoint. Stats include pod-level and node-level metrics.

 kubeletstats: auth_type: serviceAccount # Authentication mechanism with the Kubelet endpoint, refer to: https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/receiver/kubeletstatsreceiver#configuration collection_interval: 20s endpoint: ${env:OTEL_K8S_NODE_NAME}:10250 node: '${env:OTEL_K8S_NODE_NAME}' # Required to work for all CSPs without an issue insecure_skip_verify: true k8s_api_config: auth_type: serviceAccount metrics: k8s.pod.memory.node.utilization: enabled: true k8s.pod.cpu.node.utilization: enabled: true k8s.container.cpu_limit_utilization: enabled: true k8s.pod.cpu_limit_utilization: enabled: true k8s.container.cpu_request_utilization: enabled: true k8s.container.memory_limit_utilization: enabled: true k8s.pod.memory_limit_utilization: enabled: true k8s.container.memory_request_utilization: enabled: true k8s.node.uptime: enabled: true k8s.node.cpu.usage: enabled: true k8s.pod.cpu.usage: enabled: true extra_metadata_labels: - container.id 

To capture stats from every node in the cluster, deploy the Collector with the kubeletstats receiver as a DaemonSet.

The k8sclusterreceiver gathers metrics and entity events directly from the Kubernetes API server. It captures cluster-wide resources like nodes, deployments, pods, and more.

 k8s_cluster: auth_type: serviceAccount # Determines how to authenticate to the K8s API server. This can be one of none (for no auth), serviceAccount (to use the standard service account token provided to the agent pod), or kubeConfig to use credentials from ~/.kube/config. node_conditions_to_report: - Ready - MemoryPressure allocatable_types_to_report: - cpu - memory metrics: k8s.pod.status_reason: enabled: true resource_attributes: k8s.kubelet.version: enabled: true os.description: enabled: true os.type: enabled: true k8s.container.status.last_terminated_reason: enabled: true 

Run a single instance of this receiver, for example as a Deployment, with sufficient permissions to access the K8s API server.

The EDOT Collector supports a wide range of metrics receivers for popular software systems, including:

• Redis (redisreceiver): Retrieve Redis INFO data from a single Redis instance.

• JMX-based applications (jmxreceiver): Open a child Java process running the JMX Metric Gatherer configured with your specified JMX connection information and target Groovy script. It then reports metrics to an implicitly created OTLP receiver.

• Prometheus scrape targets (prometheusreceiver): Receives metric data in Prometheus format.

• Kafka (kafkareceiver): Receives telemetry data from Kafka, with configurable topics and encodings.

For a full list of supported receivers, see the EDOT Collector components reference.