Agent v2.3.0
2 minute read
July 14, 2025
New Features
- HTTP Pull/Output OAuth2.0 Support: Added support for OAuth2.0 Client Credentials flow to the HTTP Pull source and HTTP Destination nodes, enabling secure authentication with external services and improved connection security.
- Telemetry Generator Source Released: The Telemetry Generator Source node is released and the previous Demo Source and Demo Template Source nodes are deprecated.
- Distinct Count Aggregation in Aggregate Metrics: Added a distinct count aggregation method to the aggregate metric processor, supporting approximate mode (HyperLogLog) and group by key configuration for advanced metric analytics. Added an option to the aggregate metric processor to keep only fields used in group by, enabling privacy and data minimization in aggregations.
- Rate Limiter for Non-Push Sources: Added a rate limiter component to non-push based source nodes to better control data ingestion rates.
- KSM Metric Filtering and New Metric: Added support for filtering Kubernetes State Metrics (KSM) via configuration and included a new
kube_pod_status_reasonmetric, improving Kubernetes observability.
Improvements
- Supported Environments and Fleets for Source Nodes: Updated supported environments, fleets, and subfleet types, improving pipeline compatibility checking and avoiding misconfiguration.
- Refactor for GCS Destination: Updated GCS destination to no longer check agent version before requiring HMAC keys or credentials path, simplifying user configuration and avoiding unnecessary incompatibilities. Updated GCS archive strategy to take credentials definition directly and improved flexibility for various authorization mechanisms.
- Component Tag & Metrics for Tailers: Added missing component keys to certain tailers and made obsReport a no-op for component health nodes, improving telemetry labeling and reducing noise.
Bug Fixes
- Filter Empty Condition: Fixed logic in the filter processor to correctly allow all items if the context condition is empty. Fixed an issue where OTTL expressions would return errors if top-level fields were undefined, allowing valid fallback behavior.
- Remove Unused
__logical_srcField: Removed the unused field from pipeline items, reducing data size and eliminating unnecessary fields.