/nativelink

Bazel RBE with CAS server implementation in Rust. The free cache and remote execution service, prioritizing stability and speed for the people that need it.

Primary LanguageRustApache License 2.0Apache-2.0

NativeLink

License OpenSSF Scorecard OpenSSF Best Practices Slack

NativeLink is an extremely (blazingly?) fast and efficient build cache and remote executor for systems that communicate using the Remote execution protocol such as Bazel, Buck2, Goma and Reclient. NativeLink powers over one billion requests per month for customers using the system for their production workloads.

Supports Unix-based operating systems and Windows.

Getting Started with NativeLink

Below, you will find a few different options for getting started with NativeLink.

📝 Clone the NativeLink repository

  1. Go to the NativeLink repository on GitHub. Clone the repository via SSH or HTTPS. In this example the repository is cloned via SSH:
git clone git@github.com:TraceMachina/nativelink.git

📦 Installing with Cargo

  1. First install Rust, but skip to step 2 if you have it already.
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
  1. Install NativeLink with Cargo.
cargo install --git https://github.com/TraceMachina/nativelink --tag v0.4.0

⚙️ Configure and 🦾 Start NativeLink

The nativelink executable reads a JSON file as it's only parameter, --config. See nativelink-config for more details and examples.

To grab the example in your current working directory, run:

curl -O https://raw.githubusercontent.com/TraceMachina/nativelink/main/nativelink-config/examples/basic_cas.json

### you can modify the example above to replace the filesystem store with the memory store if you favor speed over data durability.
nativelink basic_cas.json

🧪 Evaluating NativeLink

  1. Once you've built NativeLink and have an instance running with the basic_cas.json configuration, launch a separate terminal session.
  2. Navigate to where you cloned the NativeLink repository:
cd $HOME/nativelink
  1. In the new terminal, run the following command to connect the running server launched above to Bazel or another RBE client:
bazel test //... \
  --remote_instance_name=main \
  --remote_cache=grpc://127.0.0.1:50051 \
  --remote_executor=grpc://127.0.0.1:50051 \
  --remote_default_exec_properties=cpu_count=1

For Windows PowerShell;

bazel test //... `
  --remote_instance_name=main `
  --remote_cache=grpc://127.0.0.1:50051 `
  --remote_executor=grpc://127.0.0.1:50051 `
  --remote_default_exec_properties=cpu_count=1

This causes Bazel to run the commands through an all-in-one CAS, scheduler and worker.

Warning

If you're using MacOS, encountering errors is anticipated at this stage. Our team is actively working on enhancing support for executing remoteable Bazel builds with MacOS. For now, you can run with Docker or a Linux virtual machine. If you have any questions, reach out on the NativeLink slack.

How it Works

This diagram is a high-level overview of the data flow in the NativeLink system. It refers to NativeLink concepts like Scheduler pool, Worker pool, and CAS rather than the cloud concepts like functions, compute nodes, and object storage to which they correspond.

sequenceDiagram
    participant build server (client)
    participant scheduler pool
    participant worker pool
    participant cas
    build server (client)->>scheduler pool: queue jobs
    scheduler pool->>worker pool: route jobs
    worker pool->>cas: upload artifacts
    worker pool->>scheduler pool: result download instructions
    scheduler pool->>build server (client): result download instructions
    cas->>build server (client): service queries
    build server (client)->>cas: service queries
Loading

❄️ Installing with Nix

Installation requirements:

This build doesn't require cloning the repository, but you need to provide a configuration file, for instance the one at nativelink-config/examples/basic_cas.json.

The following command builds and runs NativeLink in release (optimized) mode:

nix run github:TraceMachina/nativelink ./basic_cas.json

For use in production pin the executable to a specific revision:

nix run github:TraceMachina/nativelink/<revision> ./basic_cas.json

🌱 Building with Bazel

Build requirements:

  • Bazel 7.0.2
  • A recent C++ toolchain with LLD as linker

Tip

This build supports Nix/direnv which provides Bazel but no C++ toolchain (yet).

The following commands places an executable in ./bazel-bin/nativelink and starts the service:

# Unoptimized development build on Unix
bazel run nativelink -- $(pwd)/nativelink-config/examples/basic_cas.json

# Optimized release build on Unix
bazel run -c opt nativelink -- $(pwd)/nativelink-config/examples/basic_cas.json

# Unoptimized development build on Windows
bazel run --config=windows nativelink -- $(pwd)/nativelink-config/examples/basic_cas.json

# Optimized release build on Windows
bazel run --config=windows -c opt nativelink -- $(pwd)/nativelink-config/examples/basic_cas.json

Warning

The Rust compiler rustc generates numerous artifacts during compilation, including dependencies, macros, and intermediate files. When compiling programs from source, be mindful of the associated files' impact on your disk usage in the bazel-bin/ directory. This directory can grow substantially in size.

If the facing issues due to this, run the following command to clear cache files:

bazel clean --expunge

📦 Building with Cargo

Build requirements:

  • Cargo 1.74.0+
  • A recent C++ toolchain with LLD as linker

Tip

This build supports Nix/direnv which provides Cargo but no C++ toolchain/stdenv (yet).

# Unoptimized development build
cargo run --bin nativelink -- ./nativelink-config/examples/basic_cas.json

# Optimized release build
cargo run --release --bin nativelink -- ./nativelink-config/examples/basic_cas.json

Warning

The Rust compiler rustc generates numerous artifacts during compilation, including dependencies, macros, and intermediate files. When compiling programs from source, be mindful of the associated files' impact on your disk usage in the target/ directory. This directory can grow substantially in size.

If the facing issues due to this, run the following command to clear cache files:

cargo clean

🚀 Example Deployments

You can find a few example deployments in the deployment-examples directory.

See the terraform deployments for example deployments that show off remote execution and cache capabilities.

🏺 History

This project was first created due to frustration with similar projects not working or being extremely inefficient. Rust was chosen as the language to write it in because at the time Rust was going through a revolution in the new-ish feature async-await. This made making multi-threading simpler when paired with a runtime like Tokio while still giving all the lifetime and other protections that Rust gives. This pretty much guarantees that we will never have crashes due to race conditions. This kind of project seemed perfect, since there is so much asynchronous activity happening and running them on different threads is most preferable. Other languages like Go are good candidates, but other similar projects rely heavily on channels and mutex locks which are cumbersome and have to be carefully designed by the developer. Rust doesn't have these issues, since the compiler will always tell you when the code you are writing might introduce undefined behavior. The last major reason is because Rust is extremely fast and has no garbage collection (like C++, but unlike Java, Go, or Typescript).

📜 License

Copyright 2020–2024 Trace Machina, Inc.

Licensed under the Apache 2.0 License, SPDX identifier Apache-2.0.