NativeLink v0.5.0

This release implements several major overhauls: A new scheduler, better metrics and new docs.

Most changes are non-breaking. However, you might want to revisit your existing deployments since you no longer need to bundle NativeLink in remote execution worker images.

🐋 Images

docker pull ghcr.io/tracemachina/nativelink:v0.5.0

# New! This image copies nativelink to a specified location such as a mounted
# volume. You can then re-mount that volume into your worker container.
docker pull ghcr.io/tracemachina/nativelink-worker-init:v0.5.0

# The nativelink-worker-lre-cc remote execution container for Bazel with a
# Clang/LLVM 18 C++ toolchain:
docker pull ghcr.io/tracemachina/nativelink-worker-lre-cc:v0.5.0

See the contribution docs for instructions on how to build from source with Bazel, Cargo, and Nix.

📜 Changelog

❌️ Breaking changes

• The VerifyStore now auto-detects the digest function. To migrate, remove hash_verification_function settings on your config. You can toggle whether hashes should be verified with the new verify_hash boolean setting (#899).

✨ Major changes

• The scheduler internals have been rewritten. This is an important step towards high availability deployments (#1169).
• Quantities in the configs now support human readable strings such as 10s and 15KiB (#891).
• NativeLink now has experimental support for the Build Event Protocol (BEP) via the experimental_bep config option (#961).
• Worker images no longer need to bundle the nativelink executable (#1013). See the chromium example which shows how to levarage this via initContainers in K8s.
• The documentation infrastructure has been rewritten (#1056) and doc contents were overhauled (#1138). A continuously updated configuration reference is available in the docs.
• Metrics are now much more fine-grained and flexible (#1192).

See CHANGELOG.md for the full list of changes.

❤️ Thanks to our Contributors

This release contains contributions from @SchahinRohani, @aleksdmladenovic, @Abhinav-Chdhary, @eltociear, @ibilalkayy, @matdexir and various contributors at @TraceMachina.