What vectordotdev/vector shipped
Written by FoxPlug from the repository's own releases; not affiliated with Vector. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/vectordotdev/vector; Vector did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of October 5, 2026
What shipped
- Vdev release github now builds the GitHub release description from the release CUE file, including contributor and change counts, per-type counts, the upgrade guide and the release calendar, with a dry-run preview. Pull request #26584
- Schema definition caches for
possible_definitionsandinput_definitionsare now populated, cutting repeated upstream traversal when transform paths reconverge at startup. Pull request #26587 - The Debian package now ships /etc/vector/vector.yaml as a dpkg conffile so local edits survive package upgrades. Pull request #26600
- The VRL playground program editor gained completion, signature help and hover documentation. Pull request #26545
- Website browser tests were made less flaky by raising the WebKit timeout and now skip when only CUE files change. Pull request #26601
- Statically known UTF-8 values are now kept in Value::String, with source metadata, formatted values and affected tests updated. Pull request #26608
- Unix socket and framestream peer addresses plus generated syslog and DNSTAP host metadata are now kept in Value::String. Pull request #26609
- A validated ASCII character type now backs ASCII configuration fields so the Rust type, serialization and generated schema describe the same contract. Pull request #26632
- Release changelog headings use stable anchors such as breaking-changes instead of count-based ones, so section links can be built ahead of time. Pull request #26583
- GitHub release notes are rendered from a Handlebars template that can be overridden with a flag or environment variable instead of hardcoded Rust. Pull request #26619
Why it matters
This week the release tooling produces GitHub release descriptions straight from the release data and renders notes from an editable template, which makes releases easier to prepare. Debian users get /etc/vector/vector.yaml handled as a conffile, so upgrades no longer threaten local configuration. The VRL playground became easier to write expressions in, and schema definition caches now avoid repeated upstream traversal on startup.
Changelog entry
- Vdev release github now generates the GitHub release description from the release file, with counts, upgrade guide and calendar, and supports a dry run. Pull request #26584
- Upstream schema definition caches are now populated, reducing repeated traversal when transform paths reconverge at startup. Pull request #26587
- The Debian package now ships /etc/vector/vector.yaml as a conffile so local changes survive upgrades. Pull request #26600
- The VRL playground program editor now offers completion, signature help and hover documentation. Pull request #26545
- Website browser tests are less flaky on slow runners and are skipped when only CUE files change. Pull request #26601
- Statically known UTF-8 values are kept as strings in source metadata and formatted values. Pull request #26608
- Unix socket and framestream peer addresses and generated syslog and DNSTAP host metadata are kept as strings. Pull request #26609
- ASCII configuration fields use a validated character type that matches serialization and generated schema. Pull request #26632
- Release changelog headings use stable anchors so section links can be prepared in advance. Pull request #26583
- GitHub release notes render from a template that can be overridden by flag or environment variable. Pull request #26619
- The managing schemas guide examples were updated to run on current Vector. Pull request #26616
- The vdev release workflow no longer races across matrix jobs, avoiding duplicate notifications and empty draft releases. Pull request #26631
Release notes are now generated from release data with a dry-run preview, GitHub notes render from an editable template, and the Debian package ships /etc/vector/vector.yaml as a conffile. Playground editor gains completion, signature help and hover.
This week in Vector: the release tooling can build GitHub release descriptions from the release file, including contributor and change counts and a dry-run preview, and release notes now render from a template that can be overridden without recompiling. The Debian package ships /etc/vector/vector.yaml as a dpkg conffile, so local configuration survives package upgrades. The VRL playground editor gained completion, signature help and hover documentation. Schema definition caches are now populated, which avoids repeated upstream traversal when transform paths reconverge at startup.
Week of September 28, 2026
What shipped
- Datadog logs sink now optionally truncates oversized log events, preserving standard fields while shortening messages to fit configured payload limits. Pull request #26512
- OpenTelemetry source adds bounded request admission to limit concurrent requests and apply request timeouts on both HTTP and gRPC servers. Pull request #26457
- Kubernetes_logs source now caches per-pod event counter handles instead of rebuilding them for every log line, reducing memory allocation. Pull request #26490
- Unix socket configuration is now platform agnostic—types are available on all targets with platform-specific behavior isolated to the build step. Pull request #26327
- Typed trace event model added to support OTLP and Datadog trace fields without map-based storage. Pull request #26436
- Homebrew releases are now automated to stay aligned with published stable Vector versions. Pull request #26515
- OpenTelemetry protobuf definitions updated to version 1.4.0. Pull request #26513
- Transform context is now passed to VrlRunner constructors, enabling runner implementations to access build context information. Pull request #26507
- Vector test now reports ambiguous output identifiers as config errors instead of panicking. Pull request #26504
- Graph component assembly refactored to share construction logic across configuration maps and topology code. Pull request #26491
Why it matters
This week delivers improvements to data handling capacity and reliability: Datadog logs can now fit larger events through truncation, Kubernetes log ingestion uses less memory per pod, and OpenTelemetry sources apply admission control. Platform support expands with Unix socket configuration on all targets, and configuration errors surface more gracefully.
Changelog entry
- Datadog logs sink: added opt-in encoded-log and retained-message limits to truncate oversized events while preserving standard fields Pull request #26512
- OpenTelemetry source: added bounded request admission shared by HTTP and gRPC servers with
max_concurrent_requestsandrequest_timeout_secsPull request #26457 - Kubernetes_logs source: per-pod event counter handles now cached instead of rebuilt for every log line Pull request #26490
- Unix socket configuration now platform agnostic with error handling for unsupported targets Pull request #26327
- Typed trace event model added for OTLP and Datadog trace fields Pull request #26436
- Homebrew releases automated to align with stable Vector versions Pull request #26515
- OpenTelemetry protobuf definitions updated to 1.4.0 Pull request #26513
- Transform context now passed to VrlRunner constructors Pull request #26507
- Vector test now reports ambiguous output identifiers as config errors Pull request #26504
- Graph component assembly refactored to share construction logic Pull request #26491
Vector shipped: Datadog logs truncation for oversized events, Kubernetes memory optimization, OTel request admission control, Unix socket portability, and better test error reporting.
Vector's latest week brings production-focused improvements: oversized Datadog logs now truncate to fit payload limits, Kubernetes log sources cache per-pod counters for lower memory usage, and OpenTelemetry sources apply bounded request admission. Unix socket configuration works across all platforms, with error handling during tests now reporting ambiguities as config issues. Several crates continue the pedantic Clippy linting rollout.
Week of September 21, 2026
What shipped
- Datadog metrics sink now sends metrics via V3 protocol, using columnar protobuf format with dictionary-based string deduplication and delta encoding for improved efficiency. Pull request #26285
- New
files_unwatched_bytes_unread_totalmetric tracks bytes not read when files are unwatched, providing visibility into potential data loss during log rotation or file deletion. Pull request #24676 - Prometheus scrape source now supports configurable HTTP request headers, including the ability to override the default Accept header. Pull request #26075
- Datadog logs sink now allows
max_payload_bytesto be configurable. Pull request #26396 - File source now handles incomplete fingerprint headers correctly by returning UnexpectedEof so discovery can retry after the file grows. Pull request #26481
- TCP source errors now include an explicit
peer_addrfield in acknowledgement and TLS connection error logs. Pull request #26467 - Core release automation completed to remove manual merge steps and reuse published vdev binaries instead of rebuilding in each workflow. Pull request #26289
- Datadog agent and OpenTelemetry sources now write an internal %vector.trace_layout marker on emitted traces for later identification after
source_typerewriting. Pull request #26291 - Post-release housekeeping converted from auto-merged PR to direct push to master during release freeze, generating the next -dev version and restoring VRL main. Pull request #26471
- Release freeze enforcement and Kubernetes manifest refresh are now automated from configured GitHub ruleset IDs instead of hardcoded branches. Pull request #26463
Why it matters
This week brings improvements to observability metrics, sink configurations, and source capabilities while significantly advancing release automation to reduce manual developer work. Several codebase quality improvements continue the gradual rollout of pedantic linting across multiple crates.
Changelog entry
- Enhancement(datadog metrics sink): implement V3 protocol with columnar protobuf format and delta encoding Pull request #26285
- Enhancement(observability): add
files_unwatched_bytes_unread_totalmetric Pull request #24676 - Enhancement(prometheus scrape): support configurable request headers Pull request #26075
- Enhancement(datadog logs sink): make payload limit configurable via
max_payload_bytesPull request #26396 - Fix(file source): handle incomplete fingerprint headers by returning UnexpectedEof Pull request #26481
- Fix(observability): add peer addresses to TCP source error logs Pull request #26467
Vector this week: Datadog metrics V3 protocol, configurable Prometheus headers, file unwatching visibility, improved TCP error logging, and automated release workflows.
Vector ships several operator-focused improvements this week: the Datadog metrics sink now uses V3 protocol with better compression, the Prometheus scrape source supports custom headers, and file sources track unwatched bytes to surface potential data loss. Release automation advances reduce manual steps, and multiple crates graduate to stricter linting standards.
Week of September 14, 2026
What shipped
- Release preparation is now automated through vdev and CI instead of requiring maintainers to run manual Git commands locally. Pull request #26395
- The
host_metricssource now collectsmemory_oom_kill_events_totalmetric from procfs. Pull request #25802 - Release state is now managed through the root Vector package version in Cargo instead of a separate file. Pull request #26387
- JSON, Elasticsearch sink, and journald source decoding now use simdutf8 for faster UTF-8 validation with SIMD. Pull request #25610
- Vector's native event protobuf now rejects malformed data through TryFrom instead of panicking. Pull request #26292
- Systemd socket activation now correctly sets sockets to non-blocking mode for tokio compatibility. Pull request #24308
- The aggregate transform's failed updates counter is now named
aggregate_failed_updates_totalfollowing naming conventions. Pull request #26407 - Datadog protobuf generation is now extracted into lib/datadog-proto to eliminate duplicate compilation. Pull request #26422
- Databend integration tests now use RustFS instead of discontinued MinIO for S3-compatible storage. Pull request #26370
- Documentation updated to accurately reflect that VRL scripts can access the Internet and host through functions like
http_requestandget_env_var. Pull request #26434
Why it matters
Release automation improvements reduce manual maintenance burden and human error. Metric naming conventions and internal protobuf handling improvements enhance reliability. Performance optimizations through SIMD and correctness fixes in socket activation and protobuf decoding benefit all users.
Changelog entry
- Add:
host_metricssource now collectsmemory_oom_kill_events_totalmetric Pull request #25802 - Improvement: JSON, Elasticsearch, and journald decoding now use simdutf8 for faster UTF-8 validation Pull request #25610
- Fix: event protobuf now rejects malformed data with errors instead of panicking Pull request #26292
- Fix: systemd socket activation now correctly sets non-blocking mode for tokio Pull request #24308
- Fix: aggregate transform counter renamed to
aggregate_failed_updates_totalPull request #26407 - Fix:
host_metricsprocess_runtimecounter renamed toprocess_runtime_totalPull request #26406 - Fix: memory enrichment table TTL expirations now emit _total suffixed counter Pull request #26405
- Improvement: Datadog protobuf generation extracted to separate crate to eliminate duplication Pull request #26422
This week: automated release prep, faster UTF-8 validation with SIMD, OOM kill events metric, protobuf error handling that doesn't panic, and systemd socket fixes.
Vector shipped several improvements this week: release preparation is now automated to reduce manual effort, JSON and journald decoding got faster with SIMD-optimized UTF-8 validation, the host_metrics source added memory_oom_kill_events_total tracking, event protobuf handling now rejects malformed data instead of crashing, and systemd socket activation now works correctly with tokio.