[CI] Preserve Rust Docker cache across commits (#53290)

Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: OpenAI Codex <codex@openai.com>

docker/Dockerfile

@@ -253,7 +253,7 @@ ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list}
 # build stage doesn't need the rust toolchain or the rust source.
 # This stage reuses the Python environment from base and runs in parallel with
 # csrc-build/extensions-build.
-FROM base AS rust-build
+FROM base AS rust-build-cache
 ARG USE_SCCACHE
 ARG SCCACHE_ENDPOINT
 
@@ -283,26 +283,28 @@ ENV CARGO_BUILD_JOBS=4
 # separate local sccache daemon while sharing the same remote cache backend.
 ENV SCCACHE_SERVER_PORT=4227
 
-# Only Rust inputs are present, so suppress the artificial dirty state caused
-# by other tracked files being absent from this build stage.
-ENV SETUPTOOLS_SCM_PRETEND_METADATA={dirty=false}
-
-# Build the release artifacts. Cache cargo registry/git, but not target/,
-# because stale target metadata can outlive source updates across BuildKit
-# cache reuse.
+# Build with a stable version so this target snapshot is keyed only by Rust
+# inputs. The child stage below relinks the crates that embed the exact version.
 RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
     --mount=type=cache,target=/root/.cargo/git,sharing=locked \
-    --mount=type=bind,source=.git,target=.git \
     --mount=type=secret,id=aws-credentials,target=/root/.aws/credentials,required=false \
     if [ "$USE_SCCACHE" = "1" ]; then \
         if [ -n "${SCCACHE_ENDPOINT}" ]; then export SCCACHE_ENDPOINT="${SCCACHE_ENDPOINT}"; fi; \
         export RUSTC_WRAPPER=sccache; \
         sccache --show-stats; \
     fi \
-    && bash build_rust.sh \
+    && SETUPTOOLS_SCM_PRETEND_VERSION="0.0.0+docker.cache" bash build_rust.sh \
     && if [ "$USE_SCCACHE" = "1" ]; then \
         sccache --show-stats; \
     fi
+
+# Relink with the exact Git-derived package version.
+FROM rust-build-cache AS rust-build
+
+RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
+    --mount=type=cache,target=/root/.cargo/git,sharing=locked \
+    --mount=type=bind,source=.git,target=.git \
+    SETUPTOOLS_SCM_PRETEND_METADATA="{dirty=false}" bash build_rust.sh
 #################### RUST BUILD IMAGE ####################
 
 #################### CSRC BUILD IMAGE ####################

docker/Dockerfile.cpu

@@ -129,7 +129,7 @@ RUN echo 'ulimit -c 0' >> ~/.bashrc
 # Build the Rust frontend (`vllm-rs`) in a dedicated stage so the wheel build
 # stage doesn't need the rust toolchain. This stage runs in parallel
 # with the main vllm-build stage.
-FROM base AS rust-build
+FROM base AS rust-build-cache
 ENV DEBIAN_FRONTEND=noninteractive
 
 RUN apt-get update -y \
@@ -163,25 +163,27 @@ COPY build_rust.sh build_rust.sh
 # (rustc spawns enough concurrent processes to hit RLIMIT_NOFILE otherwise).
 ENV CARGO_BUILD_JOBS=4
 
-# Only Rust inputs are present, so suppress the artificial dirty state caused
-# by other tracked files being absent from this build stage.
-ENV SETUPTOOLS_SCM_PRETEND_METADATA={dirty=false}
-
-# Build the release artifacts. Cache cargo registry/git, but not target/,
-# because stale target metadata can outlive source updates across BuildKit
-# cache reuse.
+# Build with a stable version so this target snapshot is keyed only by Rust
+# inputs. The child stage below relinks the crates that embed the exact version.
 RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
     --mount=type=cache,target=/root/.cargo/git,sharing=locked \
-    --mount=type=bind,source=.git,target=.git \
     --mount=type=secret,id=aws-credentials,target=/root/.aws/credentials,required=false \
     if [ "$USE_SCCACHE" = "1" ]; then \
         export RUSTC_WRAPPER=sccache; \
         if [ -n "${SCCACHE_ENDPOINT}" ]; then export SCCACHE_ENDPOINT="${SCCACHE_ENDPOINT}"; fi; \
         sccache --show-stats; \
     fi && \
-    bash build_rust.sh && \
+    SETUPTOOLS_SCM_PRETEND_VERSION="0.0.0+docker.cache" bash build_rust.sh && \
     if [ "$USE_SCCACHE" = "1" ]; then sccache --show-stats; fi
 
+# Relink with the exact Git-derived package version.
+FROM rust-build-cache AS rust-build
+
+RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
+    --mount=type=cache,target=/root/.cargo/git,sharing=locked \
+    --mount=type=bind,source=.git,target=.git \
+    SETUPTOOLS_SCM_PRETEND_METADATA="{dirty=false}" bash build_rust.sh
+
 ######################### SOURCE PREP IMAGE #########################
 # Shared prep (source, rust artifacts, build deps) for both vllm-build and
 # vllm-dev, so the two independent compiles (bdist_wheel vs. setup.py

docker/Dockerfile.xpu

@@ -1,7 +1,7 @@
 ######################### RUST BUILD IMAGE #########################
 # Build the Rust frontend (`vllm-rs`) in a dedicated stage so the main image
 # doesn't need the rust toolchain. Runs in parallel with vllm-base.
-FROM ubuntu:22.04 AS rust-build
+FROM ubuntu:22.04 AS rust-build-cache
 ENV DEBIAN_FRONTEND=noninteractive
 
 RUN apt-get update -y \
@@ -25,14 +25,19 @@ COPY build_rust.sh build_rust.sh
 # (rustc spawns enough concurrent processes to hit RLIMIT_NOFILE otherwise).
 ENV CARGO_BUILD_JOBS=4
 
-# Only Rust inputs are present, so suppress the artificial dirty state caused
-# by other tracked files being absent from this build stage.
-ENV SETUPTOOLS_SCM_PRETEND_METADATA={dirty=false}
+# Build with a stable version so this target snapshot is keyed only by Rust
+# inputs. The child stage below relinks the crates that embed the exact version.
+RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
+    --mount=type=cache,target=/root/.cargo/git,sharing=locked \
+    SETUPTOOLS_SCM_PRETEND_VERSION="0.0.0+docker.cache" bash build_rust.sh
+
+# Relink with the exact Git-derived package version.
+FROM rust-build-cache AS rust-build
 
 RUN --mount=type=cache,target=/root/.cargo/registry,sharing=locked \
     --mount=type=cache,target=/root/.cargo/git,sharing=locked \
     --mount=type=bind,source=.git,target=.git \
-    bash build_rust.sh
+    SETUPTOOLS_SCM_PRETEND_METADATA="{dirty=false}" bash build_rust.sh
 
 FROM ubuntu:24.04 AS vllm-base
 

tests/tools/test_docker_build_metadata_args.py

@@ -153,6 +153,22 @@ def test_vllm_openai_image_embeds_metadata_contract() -> None:
         assert expected in dockerfile
 
 
+def test_rust_build_cache_excludes_git_metadata() -> None:
+    for name in ("Dockerfile", "Dockerfile.cpu", "Dockerfile.xpu"):
+        dockerfile = (REPO_ROOT / "docker" / name).read_text()
+        cached_stage, exact_version_stage = dockerfile.split(
+            "FROM rust-build-cache AS rust-build", maxsplit=1
+        )
+        exact_version_stage = exact_version_stage.split("\nFROM ", maxsplit=1)[0]
+        cached_run = cached_stage.rsplit("RUN ", maxsplit=1)[1]
+
+        assert 'SETUPTOOLS_SCM_PRETEND_VERSION="0.0.0+docker.cache"' in cached_run
+        assert "source=.git,target=.git" not in cached_run
+        assert "source=.git,target=.git" in exact_version_stage
+        assert 'SETUPTOOLS_SCM_PRETEND_METADATA="{dirty=false}"' in exact_version_stage
+        assert "bash build_rust.sh" in exact_version_stage
+
+
 def test_rocm_ci_base_bake_embeds_content_hash_label() -> None:
     bake_file = (REPO_ROOT / "docker" / "docker-bake-rocm.hcl").read_text()