DOCSEARCHbyAlgolia
Submit Docs

Vector Documentation | Log and Metrics Pipeline

website·Indexed Apr 13, 2026

https://vector.dev/docs/

Official documentation for Vector, a high-performance, open-source data routing engine for logs, metrics, and traces. The docs cover installation across Linux, macOS, and Windows via package managers or Docker, configuration of sources, sinks, and transforms, and the Vector Remap Language (VRL). Includes guides for debugging, schema management, AWS integration, and production architecture patterns like Agent and Aggregator roles.

Tokens
539.8K
Snippets
517
Records
7.4K
Agent score
50%

What's inside Vector

  1. Introduction of customConfig Helm option and TOML config deprecation

    Vector 0.15.0 introduces a new customConfig Helm option that allows direct YAML configuration in values.yaml without TOML conversion. The customConfig key is mutually exclusive with deprecated TOML-based configuration keys—if customConfig is set, no default configurations are rendered. The deprecated keys will not be removed before Vector reaches 1.0.

  2. What is Adaptive Request Concurrency (ARC)

    Adaptive Request Concurrency (ARC) is a Vector feature (introduced in version 0.11) that automatically optimizes HTTP concurrency limits based on real-time downstream service responses. Unlike static rate limits, ARC uses a feedback loop mechanism inspired by TCP congestion control algorithms to dynamically adjust request concurrency, eliminating the need for manual tuning. ARC is available on all HTTP-based sinks in Vector.

  3. What is Vector Remap Language (VRL)

    VRL is an expression-oriented language designed for transforming observability data (logs and metrics) safely and performantly. It features a simple syntax, built-in functions tailored to observability use cases, and intentionally lacks general-purpose programming constructs like loops, classes, modules, custom functions, and IO access. This simplicity ensures sound design and prevents users from writing slow or error-prone programs. VRL reached general availability (GA) in Vector 0.12.

  4. End-to-end acknowledgements planned for future release

    Vector is working on end-to-end acknowledgements to provide improved delivery guarantees. Source components will only acknowledge incoming data after that data has been successfully sent by all associated sinks (e.g., Kafka messages will not be acknowledged until downstream sinks have processed them).

  5. Known issue: Darwin tar.gz wrong architecture

    The vector-0.43.0-x86_64-apple-darwin.tar.gz executable has the wrong architecture (issue #22129). This will be fixed in v0.44. Use alternative download methods if affected.

  6. Inspect events flowing through Vector topologies with vector tap

    The vector tap subcommand (introduced in Vector 0.13) allows tapping into events flowing through Vector pipelines. Use it to live tail events during incident troubleshooting or to debug Vector configurations without routing events to a console sink.

  7. Version 0.25.1 patch release

    Patch release with fixes for regressions in 0.25.0. Users upgrading from v0.24.X should also review v0.25.0 release notes and upgrade guide.

  8. AWS SQS sink overview

    The AWS SQS sink publishes observability events to Amazon Simple Queue Service. It supports at-least-once delivery semantics, end-to-end acknowledgements, and accepts log inputs. Configure via type: "aws_sqs" with a required queue_url parameter.

  9. Vector Helm chart hosting migration

    Vector Helm charts migrated from AWS S3 bucket hosting (packages.timber.io/helm) to GitHub Pages at helm.vector.dev. The source repository is vectordotdev/helm-charts. Vector 0.16.x is the last version publishing charts to the old packages.timber.io/helm/latest and packages.timber.io/helm/nightly repositories. The new repository contains all released charts from the previous 'latest' repository.

  10. Breaking change: encoding option required for console and file sinks

    The encoding option is now required for console and file sinks. Ensure all sink configurations specify an encoding option when upgrading from earlier versions.

  11. native and native_json codecs overview

    Vector provides two experimental codecs for optimized data transmission: native (Protocol Buffers-based binary encoding) and native_json (JSON-based encoding). These codecs allow Vector to receive data directly in its native format, eliminating the need for intermediate transforms like log_to_metric. The native codec uses the same protobuf schema as the vector source and sink (defined in lib/vector-core/proto/event.proto). The native_json codec uses a JSON schema (subject to change, defined in lib/codecs/tests/data/native_encoding/schema.cue).

  12. Graceful exit when sources finish

    Vector 0.10.0 and later can gracefully exit when all configured sources finish processing. This enables Vector to be used in UNIX pipelines for batch or finite data processing tasks, replacing tools like awk and sed. The runtime monitors source completion and terminates the process automatically without manual intervention.