feat(common): implementing cargo-chef in dockerfile for better dependencies layer caching (#2228)

* feat(common): implementing cargo-chef fixing docker layer caching

* fix(common): fixed typo

* docs(common): refactor comments for better documentation

* fix(common): moved rustup target to base image, aligned utilities.dockerfile with recent changes

* fix(common): moved rustup target to base image, aligned utilities.dockerfile with recent changes

* docs(common): added doc for x86_64-unknown-linux-musl

Author: idan-starkware

Cargo.toml

@@ -24,7 +24,6 @@ members = [
     "crates/sequencing/papyrus_consensus",
 ]
 
-
 # Sequencing crates are not part of the default members.
 default-members = [
     "crates/papyrus_base_layer",

Dockerfile

@@ -1,97 +1,85 @@
-# We split the Dockerfile into two stages:
-# Stage 1: We copy all the Cargo.toml files and create empty lib.rs files.
-# Stage 2:
-#   * We copy the files from the first stage
-#   * We compile all the crates.
-#   * We copy the rest of the files and compile again.
-# The reason we compile twice is to allow caching for the first compilation (that compiles all the
-# dependency crates) if no Cargo.toml files were changed.
-# The reason we split it into two stages is to first copy all the files and then erase all
-# non-Cargo.toml files. This way, non-Cargo.toml files won't affect the cache of the second stage
-# (For more on docker stages, read https://docs.docker.com/build/building/multi-stage/).
-FROM rust:1.78 AS copy_toml
+# Dockerfile with multi-stage builds for efficient dependency caching and lightweight final image.
+# For more on Docker stages, visit: https://docs.docker.com/build/building/multi-stage/
 
-COPY crates/ /app/crates/
-COPY Cargo.toml /app/
-
-WORKDIR /app/
-
-# Erase all non-Cargo.toml files.
-RUN find /app \! -name "Cargo.toml" -type f -delete ; \
-    find /app -empty -type d -delete; \
-    # Create empty lib.rs files.
-    # In order for cargo init to work, we need to not have a Cargo.toml file. In each crate, we rename
-    # Cargo.toml to another name and after running `cargo init` we override the auto-generated
-    # Cargo.toml with the original.
-    mv Cargo.toml _Cargo.toml && \
-    # TODO: Consider moving to a script.
-    for dir in crates/*; do \
-        if [ -f "$dir/Cargo.toml" ]; then \
-            mv $dir/Cargo.toml $dir/_Cargo.toml \
-            && cargo init --lib --vcs none $dir \
-            && mv -f $dir/_Cargo.toml $dir/Cargo.toml; \
-        else \
-            for subdir in $dir/*; do \
-                mv $subdir/Cargo.toml $subdir/_Cargo.toml \
-                && cargo init --lib --vcs none $subdir \
-                && mv -f $subdir/_Cargo.toml $subdir/Cargo.toml; \
-            done; \
-        fi; \
-    done && \
-    mv _Cargo.toml Cargo.toml
-
-COPY Cargo.lock /app/
-
-# Starting a new stage so that the first build layer will be cached if a non-Cargo.toml file was
-# changed.
-# Use this image to compile the project to an alpine compatible binary.
-FROM clux/muslrust:1.78.0-stable AS builder
-WORKDIR /app/
+# We use Cargo Chef to compile dependencies before compiling the rest of the crates.
+# This approach ensures proper Docker caching, where dependency layers are cached until a dependency changes.
+# Code changes in our crates won't affect these cached layers, making the build process more efficient.
+# More info on Cargo Chef: https://github.com/LukeMathWalker/cargo-chef
 
+# We start by creating a base image using 'clux/muslrust' with additional required tools.
+FROM clux/muslrust:1.78.0-stable AS chef
+WORKDIR /app
 RUN apt update && apt install -y clang unzip
+RUN cargo install cargo-chef
 ENV PROTOC_VERSION=25.1
 RUN curl -L "https://github.com/protocolbuffers/protobuf/releases/download/v$PROTOC_VERSION/protoc-$PROTOC_VERSION-linux-x86_64.zip" -o protoc.zip && unzip ./protoc.zip -d $HOME/.local &&  rm ./protoc.zip
 ENV PROTOC=/root/.local/bin/protoc
-
-
-# Copy all the files from the previous stage (which are Cargo.toml and empty lib.rs files).
-COPY --from=copy_toml /app .
-
-# Add the proc_macro code since cargo init puts autogenerated code in lib.rs file that breaks the build.
-COPY crates/papyrus_proc_macros /app/crates/papyrus_proc_macros
-
-RUN rustup target add x86_64-unknown-linux-musl && \
-    CARGO_INCREMENTAL=0 cargo build  --target x86_64-unknown-linux-musl --release --package papyrus_node
-
-# Copy the rest of the files.
-COPY crates/ /app/crates
-
-# Touching the lib.rs files to mark them for re-compilation. Then re-compile now that all the source
-# code is available
-RUN touch crates/*/src/lib.rs; \
-    CARGO_INCREMENTAL=0 cargo build --release --package papyrus_node --bin papyrus_node;
-
-# Starting a new stage so that the final image will contain only the executable.
+# Add the x86_64-unknown-linux-musl target to rustup for compiling statically linked binaries.
+# This enables the creation of fully self-contained binaries that do not depend on the system's dynamic libraries,
+# resulting in more portable executables that can run on any Linux distribution.
+RUN rustup target add x86_64-unknown-linux-musl
+
+#####################
+# Stage 1 (planer): #
+##################### 
+FROM chef AS planner
+COPY . .
+# * Running Cargo Chef prepare that will generate recipe.json which will be used in the next stage.
+RUN cargo chef prepare
+
+#####################
+# Stage 2 (cacher): #
+##################### 
+# Compile all the dependecies using Cargo Chef cook.
+FROM chef AS cacher
+
+# Copy recipe.json from planner stage
+COPY --from=planner /app/recipe.json recipe.json
+
+# Build dependencies - this is the caching Docker layer!
+RUN cargo chef cook --target x86_64-unknown-linux-musl --release --package papyrus_node
+
+######################
+# Stage 3 (builder): #
+###################### 
+FROM chef AS builder
+COPY . .
+COPY --from=cacher /app/target target
+# Disable incremental compilation for a cleaner build.
+ENV CARGO_INCREMENTAL=0
+
+# Compile the papyrus_node crate for the x86_64-unknown-linux-musl target in release mode, ensuring dependencies are locked.
+RUN cargo build --target x86_64-unknown-linux-musl --release --package papyrus_node --locked
+
+###########################
+# Stage 4 (papyrus_node): #
+###########################
+# Uses Alpine Linux to run a lightweight and secure container.
 FROM alpine:3.17.0 AS papyrus_node
 ENV ID=1000
-
 WORKDIR /app
-# Copy the node executable and its config.
+
+# Copy the node executable and its configuration.
 COPY --from=builder /app/target/x86_64-unknown-linux-musl/release/papyrus_node /app/target/release/papyrus_node
-COPY config/ /app/config
+COPY config config
 
+# Install tini, a lightweight init system, to call our executable.
 RUN set -ex; \
     apk update; \
     apk add --no-cache tini; \
     mkdir data
 
+# Create a new user "papyrus".
 RUN set -ex; \
     addgroup --gid ${ID} papyrus; \
     adduser --ingroup $(getent group ${ID} | cut -d: -f1) --uid ${ID} --gecos "" --disabled-password --home /app papyrus; \
     chown -R papyrus:papyrus /app
 
+# Expose RPC and monitoring ports.
 EXPOSE 8080 8081
 
+# Switch to the new user.
 USER ${ID}
 
+# Set the entrypoint to use tini to manage the process.
 ENTRYPOINT ["/sbin/tini", "--", "/app/target/release/papyrus_node"]

papyrus_utilities.Dockerfile

@@ -8,22 +8,20 @@
 
 INCLUDE Dockerfile
 
-
 # Build papyrus utilities.
 FROM builder AS utilities_builder
 
 # Build papyrus_load_test and copy its resources.
-RUN CARGO_INCREMENTAL=0 cargo build --target x86_64-unknown-linux-musl --release --package papyrus_load_test --bin papyrus_load_test
+RUN cargo build --target x86_64-unknown-linux-musl --release --package papyrus_load_test --bin papyrus_load_test
 
 # Build dump_declared_classes.
-RUN CARGO_INCREMENTAL=0 cargo build --target x86_64-unknown-linux-musl --release --package papyrus_storage --features "clap" \
+RUN cargo build --target x86_64-unknown-linux-musl --release --package papyrus_storage --features "clap" \
     --bin dump_declared_classes
 
 # Build storage_benchmark.
-RUN CARGO_INCREMENTAL=0 cargo build --target x86_64-unknown-linux-musl --release --package papyrus_storage \
+RUN cargo build --target x86_64-unknown-linux-musl --release --package papyrus_storage \
     --features "clap statistical" --bin storage_benchmark
 
-
 # Starting a new stage so that the final image will contain only the executables.
 FROM alpine:3.17.0 AS papyrus_utilities