Vector is an open-source observability data pipeline maintained by Datadog's Community Open Source Engineering team. It collects logs and metrics from dozens of sources, transforms them with programmable steps including the Vector Remap Language, and routes them to the vendor or storage destination of your choice. Version 0.59.0 was the current release announced on the official site at the time of research.[1][2]
Quick verdict
Vector is a Rust observability data pipeline that collects, transforms, and routes logs and metrics. It ships as a single self-hosted binary that runs as an agent, sidecar, or aggregator, and its vendor-neutral sinks make it a practical replacement for Logstash-style collection tiers.[1][2]
What Vector can do
Vector collects logs and metrics through 47 documented sources, including files, journals, Kafka, Kubernetes logs, HTTP, and syslog receivers.[1][2]
Transforms such as remap with VRL, filter, route, sample, aggregate, and dedupe let operators reshape events inside the pipeline before delivery.[1]
Sinks cover Datadog, Elasticsearch, AWS S3, Kafka, Loki, ClickHouse, and many more, so one pipeline can fan out to several vendors at once.[1]
It deploys as a distributed agent, a sidecar, a centralized aggregator, or a stream-based topology, and installs from a single binary with no runtime.[1]
Disk buffers with documented delivery guarantees protect pipelines against downstream outages, a correctness area its own test harness checks head-to-head.[1][2]
Pipeline configuration is composable YAML, with TOML and JSON also supported, and the site publishes ready-made examples for common routes.[1]
Where Vector fits well
Teams consolidating collection agents can replace per-vendor shippers with one binary that serves logs and metrics together.[1][2]
Operators trying to cut observability spend can filter, sample, and reroute data before it reaches billed vendor endpoints.[1]
Its published throughput comparisons against Filebeat, FluentBit, FluentD, and Logstash make it worth evaluating where pipeline throughput is the bottleneck.[1]
Limits to weigh
Running Vector is self-hosted operations work: you own the YAML configuration, capacity sizing, and upgrades across Docker, Kubernetes, or OS packages.[1][2]
The project is MPL-2.0 licensed, so teams with strict license policies should review the file-level copyleft terms before embedding it.[1]
No first-party managed Vector Cloud was documented, so teams wanting a hosted pipeline must look to other offerings.[1]
No SSO documentation was found for the pipeline itself, so access control expectations should be verified before adopting it in regulated setups.[1][2]
Questions worth asking
Is Vector free to use?
What can Vector replace?
It covers the same log-ingest-transform-ship workflow as Logstash, Fluentd, and similar shippers, and its own repository benchmarks it directly against those tools. Migration effort depends on your existing pipeline configuration.[1]
How do you deploy Vector?
Sources and editorial review
Source research. Reviewed on October 13, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Vector official site
official · Observed October 13, 2026
Official homepage positioning, deployment roles, configuration examples, and the 0.59.0 release announcement.
- vectordotdev/vector repository
official · Observed October 13, 2026
README definition, Datadog maintenance statement, performance and correctness comparisons, and community adoption figures.
- Vector license file
official · Observed October 13, 2026
Repository license text identifying the Mozilla Public License version 2.0.
- Vector installation and component docs
official · Observed October 13, 2026
Supported package managers, platforms, and the source, transform, and sink catalog including VRL remap.
Compared with these products
No explicit proprietary baseline is recorded.