slurm-local-integration-tests.yaml

name: Slurm Integration Tests

on:
  push:
    branches:
      - main
  pull_request:

env:
  SLURM_VERSION: 21.08.6

jobs:
  slurm:
    runs-on: linux.20_04.4x
    permissions:
      id-token: write
      contents: read
    steps:
      - name: Setup Python
        uses: actions/setup-python@v2
        with:
          python-version: 3.8
          architecture: x64
      - name: Checkout TorchX
        uses: actions/checkout@v2
      - name: Checkout slurm-docker-cluster
        run: |
          set -ex

          # TODO: switch to trunk once https://github.com/giovtorres/slurm-docker-cluster/pull/29 lands
          git clone https://github.com/d4l3k/slurm-docker-cluster.git
      - name: Pull docker containers
        run: |
          set -ex

          cd slurm-docker-cluster
          docker compose pull --ignore-pull-failures
        continue-on-error: true
      # cache layers to avoid rebuilding slurm
      - uses: satackey/action-docker-layer-caching@v0.0.11
        continue-on-error: true
      - name: Build slurm
        run: |
          set -ex
          cd slurm-docker-cluster
          docker build -t slurm-docker-cluster:$SLURM_VERSION .
      - name: Start slurm
        run: |
          set -ex
          cd slurm-docker-cluster
          IMAGE_TAG=$SLURM_VERSION docker compose up --detach
      - name: Install Dependencies
        run: |
          set -ex
          pip install wheel
      - name: Run Slurm Integration Tests
        run: |
          set -e

          mkdir -p ~/.ssh
          echo "$SLURM_KNOWN_HOST" >> ~/.ssh/known_hosts

          scripts/slurmdockerint.sh