dex/Dockerfile.dvenue

# ---------------------------------------------------------------------------
# D-Chain VENUE image — ghcr.io/luxfi/dchain-venue
#
# The single-operator private build that actually serves the D-Chain. It is NOT
# one of the 5 public luxd validators and is NOT the public ghcr.io/luxfi/node
# image (which excludes the DexVM). It bundles TWO binaries the operator venue
# StatefulSet runs from one image:
#
#   /usr/local/bin/dvenue   — the ZAP matching engine (luxfi/dex cmd/dvenue,
#                             //go:build cgo, built CGO_ENABLED=1). This is the
#                             real CLOB:
#                             it links the GPU matching kernels (libdex_clob_cuda
#                             + libamm_xyk_cuda) and the secp256k1 ecrecover
#                             bundle via pkg-config. The 0x9010 C-Chain precompile
#                             and the chains/dexvm atomic proxy dial it.
#   /luxd/build/luxd        — the dchain-tagged node (`-tags dchain`), which
#                             links the chains/dexvm PROXY so the venue node
#                             serves the P-chain-registered D-Chain. The proxy
#                             relays to the dvenue socket; it is pure-Go (no GPU).
#
# ARCH: linux/arm64 ONLY. The matcher's `cgo && linux` path in luxfi/dex pkg/lx
# (orderbook_cuda.go + amm_gpu_cuda.go) has no non-CUDA fallback on any arch, so
# the venue can only be built where CUDA exists — the arcd fleet's one CUDA host
# is spark (NVIDIA GB10, sm_121, aarch64). There is no linux/amd64 CUDA runner.
#
# PRIVATE DEPS: the three pkg-config libs are fetched as a per-arch tarball from
# the luxcpp/dex `dex-gpu-release` workflow (luxdex-venue-<ver>-linux-arm64.tar.gz),
# extracted to /usr/local — exactly how node/Dockerfile fetches luxcpp/cevm.
# That tarball carries lux-dex-clob-cuda.pc, lux-dex-amm-cuda.pc and
# lux-crypto-secp256k1.pc plus their libs/headers under a GNU prefix.
# ---------------------------------------------------------------------------
ARG GO_VERSION=1.26.4
# CUDA 12.x devel image (arm64/sbsa variant) provides nvcc-free libcudart for
# the cgo link of dvenue. Runtime stage uses the matching -runtime image.
ARG CUDA_VERSION=12.4.1
ARG UBUNTU_VERSION=22.04

# ============= Builder Stage (arm64, CUDA devel) ================
# CUDA devel base gives libcudart + headers so cgo can link -lcudart. Go is
# installed on top (the project requires Go 1.26.x, newer than apt ships).
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION} AS builder

ARG GO_VERSION
RUN apt-get update && apt-get install -y --no-install-recommends \
    gcc g++ libc6-dev make git ca-certificates wget pkg-config xz-utils \
    && rm -rf /var/lib/apt/lists/*

# Install Go (arm64).
RUN wget -q "https://go.dev/dl/go${GO_VERSION}.linux-arm64.tar.gz" \
    && tar -C /usr/local -xzf "go${GO_VERSION}.linux-arm64.tar.gz" \
    && rm "go${GO_VERSION}.linux-arm64.tar.gz"
ENV PATH="/usr/local/go/bin:${PATH}"

# luxfi module resolution: skip checksum DB (tags may be rewritten), proxy for
# public deps, direct for luxfi/*.
ENV GONOSUMCHECK=github.com/luxfi/*
ENV GONOSUMDB=github.com/luxfi/*
ENV GOPROXY=https://proxy.golang.org,direct
ENV GONOPROXY=github.com/luxfi/*
ENV GOFLAGS="-mod=mod"
ENV GOEXPERIMENT=jsonv2

# Private luxfi/* modules resolve via git url.insteadOf using a BuildKit secret
# (same shape as node/Dockerfile). The token is written to /etc/gitconfig on
# this throwaway builder stage; only the compiled binaries reach the runtime.
RUN --mount=type=secret,id=ghtok,required=false \
    if [ -s /run/secrets/ghtok ]; then \
        git config --global url."https://x-access-token:$(cat /run/secrets/ghtok)@github.com/".insteadOf "https://github.com/"; \
    fi

# --- Fetch the venue GPU dependency bundle (the private link surface) --------
# Published by luxcpp/dex .github/workflows/dex-gpu-release.yml. Extracting to
# /usr/local puts the .pc files at /usr/local/lib/pkgconfig and the libs at
# /usr/local/lib, so pkg-config resolves lux-dex-clob-cuda / lux-dex-amm-cuda /
# lux-crypto-secp256k1 at the cgo link step below.
#
# CI/RELEASE GAP (until the dex-gpu-release workflow has run on a tag): the URL
# below 404s. The build then fails at the dvenue link step with a clear
# pkg-config error. Run luxcpp/dex dex-gpu-release first (or point VENUE_DEPS_URL
# at the workflow artifact) to populate it.
ARG VENUE_DEPS_VERSION=v0.1.1
ARG VENUE_DEPS_URL=https://github.com/luxcpp/dex/releases/download/${VENUE_DEPS_VERSION}/luxdex-venue-${VENUE_DEPS_VERSION}-linux-arm64.tar.gz
RUN --mount=type=secret,id=ghtok,required=false \
    AUTH=""; [ -s /run/secrets/ghtok ] && AUTH="--header=Authorization: Bearer $(cat /run/secrets/ghtok)"; \
    wget -q $AUTH "${VENUE_DEPS_URL}" -O /tmp/venue-deps.tar.gz \
    && tar -xzf /tmp/venue-deps.tar.gz -C /usr/local \
    && rm /tmp/venue-deps.tar.gz \
    && ldconfig 2>/dev/null || true
ENV PKG_CONFIG_PATH=/usr/local/lib/pkgconfig

# --- Build dvenue (the GPU matcher) from luxfi/dex ---------------------------
# cmd/dvenue is `//go:build cgo`: CGO_ENABLED=1 (which sets the `cgo` build tag)
# selects pkg/lx's GPU matcher (orderbook_cuda + amm_gpu_cuda + signed_order_gpu,
# all gated on `cgo`), satisfied by the pkg-config libs above. `-tags cgo` is
# explicit-for-readability (CGO_ENABLED=1 already implies it). The pure-Go CPU
# venue is cmd/dvenue-cpu (//go:build !cgo) — not built here.
WORKDIR /build/dex
COPY go.mod go.sum ./
RUN go mod download
COPY . .
RUN CGO_ENABLED=1 GOOS=linux GOARCH=arm64 \
    go build -trimpath -tags "cgo" -ldflags="-s -w" \
        -o /out/dvenue ./cmd/dvenue

# --- Build the dchain-tagged luxd (the proxy node) ---------------------------
# Cloned from luxfi/node at a pinned release and built `-tags dchain` so it
# links the chains/dexvm proxy (pure-Go; published chains dep). The dchain tag
# is appended to the build profile's tags so nattraversal etc. are preserved.
ARG NODE_VERSION=v1.30.12
RUN git clone --depth 1 --branch ${NODE_VERSION} https://github.com/luxfi/node.git /build/node
WORKDIR /build/node
RUN go mod download
# Tag set = the public node image's `minimal` profile (`nattraversal`) PLUS
# `dchain` so the chains/dexvm proxy links. ZAP transport is in the default
# build (not gated). Deliberately NO `grpc`: the `grpc` tag flips
# luxfi/zapdb/pb to the protobuf variant (badgerpb4.pb.go), which drops the
# package-level Marshal/Unmarshal that zapdb/table needs — `grpc` does not
# compile against zapdb v1.10.1 (verified). The venue node needs no gRPC plugin
# transport, so `minimal,dchain` is both correct and sufficient.
#
# CGO_ENABLED=0: the dchain luxd's only cgo edge is the optional luxcpp/cevm
# C-Chain backend; the dexvm proxy + pure-Go EVM need no native libs (the
# build graph pulls only runtime/cgo). Building pure-Go keeps the venue node
# free of the cevm fetch — the GPU/cgo surface lives entirely in dvenue.
RUN GITCOMMIT=$(git rev-parse HEAD) && \
    CGO_ENABLED=0 GOOS=linux GOARCH=arm64 \
    go build -trimpath -tags "nattraversal,dchain" -ldflags="-s -w \
        -X github.com/luxfi/node/version.GitCommit=${GITCOMMIT}" \
        -o /luxd/build/luxd ./main

# --- Build the EVM (C-Chain) plugin ------------------------------------------
# node registers the optional VMs (incl. the dexvm proxy) + P/X in-process, but
# the C-Chain EVM is an EXTERNAL plugin (same as the public node image's EVM
# Plugin Stage). The venue --track-all-chains, and the 0x9010 dex precompile
# (NewZAPEngine(dex-zap-endpoint) -> the dvenue socket) lives in this plugin, so
# the venue MUST ship it. Built from the pinned luxfi/evm release, CGO=0,
# written to /luxd/build/plugins/<EVM_VM_ID> — startup.sh seeds it into
# /data/plugins (--plugin-dir).
ARG EVM_VERSION=v1.99.25
ARG EVM_VM_ID=mgj786NP7uDwBCcq6YwThhaN8FLyybkCa4zBWTQbNgmK6k9A6
RUN git clone --depth 1 --branch ${EVM_VERSION} https://github.com/luxfi/evm.git /build/evm && \
    cd /build/evm && \
    mkdir -p /luxd/build/plugins && \
    CGO_ENABLED=0 GOOS=linux GOARCH=arm64 GOFLAGS=-mod=mod \
    go build -trimpath -ldflags="-s -w" -o /luxd/build/plugins/${EVM_VM_ID} ./plugin && \
    chmod +x /luxd/build/plugins/${EVM_VM_ID}

# ============= Runtime Stage (arm64, CUDA runtime) ================
# CUDA -runtime image carries libcudart for the GPU kernels at run time.
FROM nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION} AS execution

RUN apt-get update && apt-get install -y --no-install-recommends \
    curl ca-certificates \
    && rm -rf /var/lib/apt/lists/*

# Carry the GPU/crypto shared libs the dvenue binary may dlopen / rpath into.
COPY --from=builder /usr/local/lib /usr/local/lib
COPY --from=builder /usr/local/include /usr/local/include
RUN ldconfig 2>/dev/null || true

# Binaries: the dchain luxd (venue node) + the EVM (C-Chain) plugin it loads +
# dvenue (ZAP matcher). startup.sh seeds /luxd/build/plugins/* into /data/plugins.
COPY --from=builder /luxd/build/luxd /luxd/build/luxd
COPY --from=builder /luxd/build/plugins /luxd/build/plugins
COPY --from=builder /out/dvenue /usr/local/bin/dvenue

# Non-root (matches the operator StatefulSet securityContext runAsUser 1000).
RUN groupadd -g 1000 venue && useradd -u 1000 -g venue -m -s /usr/sbin/nologin venue \
    && mkdir -p /data && chown -R venue:venue /data /luxd
USER venue

ENV PATH="/luxd/build:/usr/local/bin:${PATH}"
WORKDIR /luxd/build

# 9650 luxd HTTP RPC, 9651 luxd staking, 9099 dvenue ZAP socket.
EXPOSE 9650 9651 9099

# Default to the venue node; the operator overrides the `zap` container's
# command to /usr/local/bin/dvenue.
CMD ["/luxd/build/luxd"]