helexa/.gitea/workflows/deploy-dev.yml

name: deploy-dev

# Fast-path iteration deploy for a SINGLE neuron host: build one CUDA
# flavour, copy the raw binary to the host, restart neuron.service.
# Skips the other two flavours, all RPM packaging, signing, repo
# publish, and dnf — push-to-testable drops from ~20 min to roughly
# one CUDA build plus a service restart.
#
# This is a DEV convenience, not a release path:
#   - the binary lands at /usr/bin/neuron *outside* RPM ownership;
#     the next regular deploy.yml run reconciles the host back to the
#     packaged binary (dnf sees the newer RPM and reinstalls). `rpm -V
#     helexa-neuron-<flavour>` flagging a modified /usr/bin/neuron in
#     the interim is expected.
#   - nothing is published; other hosts are untouched.
#   - requires the `install` sudoers rule from
#     asset/sudoers.d/neuron-host.conf (re-run script/infra-setup.sh
#     after updating it).
#
# Trigger from the Gitea UI: Actions → deploy-dev → Run workflow,
# pick the target host. Defaults to the ref you dispatch from, so it
# works from feature branches without touching main.

on:
  workflow_dispatch:
    inputs:
      target:
        description: "neuron host to deploy to"
        required: true
        type: choice
        options: [beast, benjy, quadbrat]
        default: beast

# One dev deploy at a time; a newer dispatch for the same host wins.
concurrency:
  group: deploy-dev-${{ inputs.target }}
  cancel-in-progress: true

env:
  CARGO_INCREMENTAL: "0"
  CARGO_TERM_COLOR: "always"

jobs:
  build:
    name: Build neuron (${{ inputs.target }})
    runs-on: cuda-13.0
    outputs:
      flavour: ${{ steps.map.outputs.flavour }}
    steps:
      - uses: actions/checkout@v4

      # host → flavour → compute cap. Keep in sync with the
      # build-neuron matrix in build-prerelease.yml and the
      # deploy-neurons matrix in deploy.yml.
      - id: map
        run: |
          case "${{ inputs.target }}" in
            beast)    flavour=blackwell cap=120 ;;
            benjy)    flavour=ada       cap=89  ;;
            quadbrat) flavour=ampere    cap=86  ;;
            *) echo "unknown target ${{ inputs.target }}"; exit 1 ;;
          esac
          echo "flavour=${flavour}" >> "$GITHUB_OUTPUT"
          echo "cap=${cap}" >> "$GITHUB_OUTPUT"

      - name: Build neuron with CUDA
        run: |
          set -eux
          export PATH="/usr/local/cuda-13.0/bin:${PATH}"
          export LD_LIBRARY_PATH="/usr/local/cuda-13.0/targets/x86_64-linux/lib:/usr/local/cuda-13.0/lib64:${LD_LIBRARY_PATH:-}"
          export LIBRARY_PATH="/usr/local/cuda-13.0/targets/x86_64-linux/lib:/usr/local/cuda-13.0/lib64:${LIBRARY_PATH:-}"
          cargo build --release -p neuron --features "cuda cudnn"
        env:
          CUDA_COMPUTE_CAP: ${{ steps.map.outputs.cap }}
          CARGO_BUILD_JOBS: "8"
          NVCC_THREADS: "4"

      - name: Stage binary
        run: |
          mkdir --parents artifacts
          cp target/release/neuron artifacts/neuron-dev
          file artifacts/neuron-dev

      - uses: actions/upload-artifact@v3
        with:
          name: neuron-dev-${{ inputs.target }}
          path: artifacts/neuron-dev
          retention-days: 1

  deploy:
    name: Deploy to ${{ inputs.target }}
    needs: build
    runs-on: fedora-43
    env:
      DEPLOY_KEY: |
        ${{ secrets.RSYNC_SSH_KEY }}
      TARGET_HOST: ${{ inputs.target }}.hanzalova.internal
    steps:
      - name: SSH init
        run: |
          mkdir -p ~/.ssh
          echo "${DEPLOY_KEY}" > ~/.ssh/id_ed25519
          chmod 600 ~/.ssh/id_ed25519
          ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=accept-new \
              "gitea_ci@${TARGET_HOST}" 'hostname -f'

      - uses: actions/download-artifact@v3
        with:
          name: neuron-dev-${{ inputs.target }}
          path: artifacts/

      - name: Copy binary to host
        run: |
          scp artifacts/neuron-dev "gitea_ci@${TARGET_HOST}:/var/lib/gitea_ci/neuron-dev"

      - name: Install binary and restart neuron.service
        run: |
          ssh "gitea_ci@${TARGET_HOST}" '
            set -eu
            if systemctl is-active --quiet neuron.service; then
              sudo /usr/bin/systemctl stop neuron.service
            fi
            # Exact command form required by the sudoers rule in
            # asset/sudoers.d/neuron-host.conf — change both together.
            sudo /usr/bin/install -o root -g root -m 0755 /var/lib/gitea_ci/neuron-dev /usr/bin/neuron
            # enable --now so a dev deploy also leaves the unit enabled
            # for boot, consistent with deploy.yml.
            sudo /usr/bin/systemctl enable --now neuron.service
            rm -f /var/lib/gitea_ci/neuron-dev'

      - name: Capture neuron.service startup journal
        if: always()
        run: |
          sleep 10
          ssh "gitea_ci@${TARGET_HOST}" \
              'journalctl --unit neuron.service -I --no-pager'