Grafana Agent

grafana/agent last check crawl paused · since 2026-09-25 99 releases
Notes

Grafana Agent has reached End-of-Life (EOL) on November 1, 2025. It is recommended migrating to the Grafana Alloy collector, which is built on the foundation of Grafana Agent Flow.

Release notes
0.39.0 · 1y+
view on github

This is release v0.39.0 of Grafana Agent.

Upgrading

Read the relevant upgrade guides for specific instructions on upgrading from older versions:

Notable changes:

Breaking changes

  • otelcol.receiver.prometheus will drop all otel_scope_info metrics when converting them to OTLP. (@wildum)

    • If the otel_scope_info metric has labels otel_scope_name and otel_scope_version, their values will be used to set OTLP Instrumentation Scope name and version respectively.
    • Labels of otel_scope_info metrics other than otel_scope_name and otel_scope_version are added as scope attributes with the matching name and version.

  • The target block in prometheus.exporter.blackbox requires a mandatory name argument instead of a block label. (@hainenber)

  • In the azure exporter, dimension options will no longer be validated by the Azure API. (@kgeckhart)

    • This change will not break any existing configurations and you can opt in to validation via the validate_dimensions configuration option.
    • Before this change, pulling metrics for azure resources with variable dimensions required one configuration per metric + dimension combination to avoid an error.
    • After this change, you can include all metrics and dimensions in a single configuration and the Azure APIs will only return dimensions which are valid for the various metrics.

Features

  • A new discovery.ovhcloud component for discovering scrape targets on OVHcloud. (@ptodev)
  • Allow specifying additional containers to run. (@juangom)

Enhancements

  • Flow Windows service: Support environment variables. (@jkroepke)

  • Allow disabling collection of root Cgroup stats in prometheus.exporter.cadvisor (flow mode) and the cadvisor integration (static mode). (@hainenber)

  • Grafana Agent on Windows now automatically restarts on failure. (@hainenber)

  • Added metrics, alerts and dashboard visualisations to help diagnose issues with unhealthy components and components that take too long to evaluate. (@thampiotr)

  • The http config block may now reference exports from any component. Previously, only remote.* and local.* components could be referenced without a circular dependency. (@rfratto)

  • Add support for Basic Auth-secured connection with Elasticsearch cluster using prometheus.exporter.elasticsearch. (@hainenber)

  • Add a resource_to_telemetry_conversion argument to otelcol.exporter.prometheus for converting resource attributes to Prometheus labels. (@hainenber)

  • pyroscope.ebpf support python on arm64 platforms. (@korniltsev)

  • otelcol.receiver.prometheus does not drop histograms without buckets anymore. (@wildum)

  • Added exemplars support to otelcol.receiver.prometheus. (@wildum)

  • mimir.rules.kubernetes may now retry its startup on failure. (@hainenber)

  • Added links between compatible components in the documentation to make it easier to discover them. (@thampiotr)

  • Allow defining HTTPClientConfig for discovery.ec2. (@cmbrad)

  • The remote.http component can optionally define a request body. (@tpaschalis)

  • Added support for loki.write to flush WAL on agent shutdown. (@thepalbi)

  • Add support for integrations-next static to flow config conversion. (@erikbaranowski)

  • Add support for passing extra arguments to the static converter such as -config.expand-env. (@erikbaranowski)

  • Added 'country' mmdb-type to log pipeline-stage geoip. (@superstes)

  • Azure exporter enhancements for flow and static mode, (@kgeckhart)

    • Allows for pulling metrics at the Azure subscription level instead of resource by resource
    • Disable dimension validation by default to reduce the number of exporter instances needed for full dimension coverage

  • Add max_cache_size to prometheus.relabel to allow configurability instead of hard coded 100,000. (@mattdurham)

  • Add support for http_sd_config within a scrape_config for prometheus to flow config conversion. (@erikbaranowski)

  • discovery.lightsail now supports additional parameters for configuring HTTP client settings. (@ptodev)

  • Add sample_age_limit to remote_write config to drop samples older than a specified duration. (@marctc)

For a full list of changes, please refer to the CHANGELOG!

Installation

Refer to our installation guides for how to install the variants of Grafana Agent: