Vector | v0.55.0

Vector v0.55.0 released on 22-04-2026

Vector v0.55.0 is out now. It introduces a native windows_event_log source, Apache Parquet output for aws_s3, restored Azure Blob authentication and datadog_metrics v2 defaults, moves the observability API from GraphQL to gRPC, and fixes a CPU regression affecting file and kubernetes_logs sources.

Read the full release notes at https://vector.dev/releases/0.55.0 and follow the 0.55 upgrade guide for migration steps: 0.55 upgrade guide.

What’s in this release

• New windows_event_log source that reads Windows Event Log channels via the Windows Event Log API with pull-mode subscriptions, bookmark-based checkpointing, and configurable field filtering.
• aws_s3 sink gains Apache Parquet support: events can be written as columnar Parquet files with an auto-generated native schema or a supplied .schema file, Parquet batching, and compression options (Snappy, ZSTD, GZIP, LZ4, or none).
• Observability API moved from GraphQL to gRPC (affects tooling that queried /graphql or used the playground), vector top accuracy fixes, new internal metrics for capacity/backpressure, task-transform utilization improvements, and a fix for a buffer utilization underflow regression.

Upgrade notes

• Breaking change: observability API has moved from GraphQL to gRPC — migrate any tooling that queried /graphql or used the playground (including vector top and vector tap). See the 0.55 upgrade guide for migration steps: upgrade guide.
• Other changes to review: removal of the top-level headers option on the http and opentelemetry sinks, and the explicit azure_credential_kind requirement for azure_logs_ingestion when using Client Secret credentials; follow the upgrade guide for recommended actions. No separate rollback procedure is documented in the release notes—refer to the guide for advice on migration and recovery.

Tell us how the upgrade goes or share any issues you hit — feedback helps the project and other users.