Single-stage Dockerfile.stage: pre-download k3s on host runner
Dev — Build & local smoke test / build-smoke (pull_request) Failing after 1m37s

The multi-stage Docker builder (which installed build-essential, compiled
soul-demo, and downloaded k3s inside Docker) was causing RWLayer nil
corruption on the runner's overlay2 driver. Every affected run failed at
apt-get install in the runtime stage after the builder stage completed.

Fix: move k3s download to the CI host runner (same pattern as soul-demo
compilation, which now passes reliably). Dockerfile.stage becomes single-
stage: no apt-get in a builder stage, no network downloads, just COPY of
pre-built binaries. Also adds --no-cache to the main docker build for
consistency with the soul-demo step fix.
This commit is contained in:
2026-05-10 11:26:23 -05:00
parent 5e1344af42
commit e6fd110073
2 changed files with 31 additions and 47 deletions
+17 -2
View File
@@ -196,13 +196,28 @@ jobs:
echo "soul-demo-image.tar: $(du -sh dist/soul-demo-image.tar | cut -f1)"
docker rmi soul-demo:local 2>/dev/null || true
- name: Download k3s binary
# Pre-download k3s on the host runner so Dockerfile.stage can COPY it
# directly. Previously k3s was downloaded inside the Docker builder stage,
# which combined with build-essential and C compilation caused RWLayer nil
# corruption on the runner's overlay2 driver. Host-runner download is safe.
if: steps.changetype.outputs.asset_only != 'true'
run: |
set -euo pipefail
curl -fL --retry 3 --retry-delay 10 \
https://github.com/k3s-io/k3s/releases/download/v1.32.4%2Bk3s1/k3s \
-o dist/k3s
chmod +x dist/k3s
echo "k3s: $(ls -lh dist/k3s)"
- name: Build and tag image
if: steps.changetype.outputs.asset_only != 'true'
run: |
set -euo pipefail
# --no-cache: prevents reuse of corrupted overlay2 layers from prior failed runs.
# Dockerfile.stage is now single-stage (no builder) so build is fast even without cache.
docker build \
--build-arg BUILDKIT_INLINE_CACHE=1 \
--cache-from us-central1-docker.pkg.dev/neuron-785695/neuron-marketing/marketing:stage-latest \
--no-cache \
-f Dockerfile.stage \
-t "marketing:${{ steps.tag.outputs.tag }}" \
.
+14 -45
View File
@@ -4,50 +4,16 @@
# - neuron-web on port 8080 (landing page server)
# - soul-demo on port 7772 (demo chat, localhost only)
#
# neuron-web is built by `elb build` in CI (not here). elb compiles each
# .el source independently and links the result — no combined mega-file,
# no exponential memory growth. The binary lands at dist/neuron-landing
# (linux/amd64) and is COPY'd directly into the runtime image.
# All binaries (neuron-web, soul-demo, k3s) are pre-built by CI on the host
# runner before this Dockerfile runs. This keeps the Docker build single-stage
# with no compilation and no network downloads, eliminating the multi-stage
# complexity that caused RWLayer corruption on the runner's overlay2 driver.
#
# soul-demo.c is pre-committed (small, no OOM risk) and compiled here.
# CI pre-build steps (in stage.yaml):
# - neuron-web: built by `elb build` → dist/neuron-landing
# - soul-demo: compiled by cc on host → dist/soul-demo
# - k3s: downloaded by curl on host → dist/k3s
# ── Stage 1: compile soul-demo ────────────────────────────────────────────────
FROM debian:bookworm-slim AS builder
RUN apt-get update \
&& apt-get install -y --no-install-recommends \
build-essential \
curl \
libcurl4-openssl-dev \
libssl-dev \
ca-certificates \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /build
COPY runtime/el_runtime.c runtime/el_runtime.h ./
# Pre-compile el_runtime as a separate cached layer.
# el_runtime.c changes rarely; main.c changes every run.
# Splitting this out means el_runtime.o is cached across builds when only main.c changes.
# -DHAVE_CURL: the staged el_runtime.c (from el.git) guards the OTLP observability
# section (emit_metric, emit_log, trace_span_*) behind #ifdef HAVE_CURL.
# libcurl IS installed above, so define HAVE_CURL to enable those functions.
RUN cc -O2 -DHAVE_CURL -c el_runtime.c -I. -o el_runtime.o
COPY dist/soul-demo.c dist/vessel_stubs.c ./
RUN cc -O2 -rdynamic \
-DEL_SOUL_DEMO_BUILD \
-o soul-demo \
soul-demo.c vessel_stubs.c el_runtime.o \
-lcurl -lpthread -ldl -lm -lssl -lcrypto
# ── Download k3s binary ───────────────────────────────────────────────────────
RUN curl -fL --retry 3 --retry-delay 10 https://github.com/k3s-io/k3s/releases/download/v1.32.4%2Bk3s1/k3s -o /usr/local/bin/k3s \
&& chmod +x /usr/local/bin/k3s
# ── Stage 2: runtime image ────────────────────────────────────────────────────
FROM debian:bookworm-slim
RUN apt-get update \
@@ -67,10 +33,13 @@ RUN apt-get update \
COPY dist/neuron-landing /usr/local/bin/neuron-web
RUN chmod +x /usr/local/bin/neuron-web
COPY --from=builder /build/soul-demo /usr/local/bin/soul-demo
# soul-demo binary — compiled by cc on host runner in CI
COPY dist/soul-demo /usr/local/bin/soul-demo
RUN chmod +x /usr/local/bin/soul-demo
# k3s binary
COPY --from=builder /usr/local/bin/k3s /usr/local/bin/k3s
# k3s binary — downloaded from GitHub releases by CI
COPY dist/k3s /usr/local/bin/k3s
RUN chmod +x /usr/local/bin/k3s
# soul-demo OCI image tar — k3s imports this at startup (no registry needed)
RUN mkdir -p /var/lib/rancher/k3s/agent/images