#283: Added GPU Support for run-db-test

.github/workflows/gpu-tests.yml

@@ -0,0 +1,60 @@
+name: GPU Tests
+
+on:
+  workflow_call:
+
+jobs:
+
+  prep-testbed-gpu:
+    runs-on: ubuntu-24.04
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Python & Poetry Environment
+        uses: exasol/python-toolbox/.github/actions/[email protected]
+
+      - id: set-matrix
+        run: |
+          ALL_TESTS=`poetry run -- nox -s integration-test-list -- --test-set gpu-only`
+          echo "matrix=$ALL_TESTS" >> "$GITHUB_OUTPUT"
+    outputs:
+      matrix: ${{ steps.set-matrix.outputs.matrix }}
+
+  test-gpu:
+    needs: prep-testbed-gpu
+    strategy:
+      fail-fast: false
+      matrix:
+        python_version:
+          - "3.10"
+        test-path: ${{fromJson(needs.prep-testbed-gpu.outputs.matrix)}}
+    runs-on:
+      labels: int-linux-x64-4core-gpu-t4-ubuntu24.04-1
+    name: GPU Test
+    steps:
+      - uses: actions/checkout@v4
+
+        #Cannot use exasol/python-toolbox/.github/actions/python-environment here
+        #because pipx is not installed on int-linux-x64-4core-gpu-t4-ubuntu24.04-1
+      - name: Setup Python ('${{ matrix.python_version }}')
+        uses: actions/setup-python@v5
+        with:
+          python-version: '${{ matrix.python_version }}'
+
+      - name: Install pipx and poetry
+        run: |
+          pip install pipx
+          pipx install poetry=="2.1.2"
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+      - name: Setup Poetry
+        run: |
+          poetry env use python3
+          poetry install
+
+      - name: Allow unprivileged user namespaces
+        run: sudo sysctl -w kernel.apparmor_restrict_unprivileged_userns=0
+
+      - name: Run ${{ matrix.test-path.name }}
+        run: poetry run -- python "${{ matrix.test-path.path }}"

.github/workflows/integration-tests.yml

@@ -0,0 +1,53 @@
+name: Integration Tests
+
+on:
+  workflow_call:
+    secrets:
+      DOCKER_USERNAME:
+          required: true
+      DOCKER_PASSWORD:
+          required: true
+jobs:
+
+  prep-testbed:
+    runs-on: ubuntu-24.04
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Python & Poetry Environment
+        uses: exasol/python-toolbox/.github/actions/[email protected]
+
+      - id: set-matrix
+        run: |
+          ALL_TESTS=`poetry run -- nox -s integration-test-list -- --test-set default`
+          echo "matrix=$ALL_TESTS" >> "$GITHUB_OUTPUT"
+    outputs:
+      matrix: ${{ steps.set-matrix.outputs.matrix }}
+
+  test-all:
+    needs: prep-testbed
+    strategy:
+      fail-fast: false
+      matrix:
+        python_version:
+          - "3.10"
+        test-path: ${{fromJson(needs.prep-testbed.outputs.matrix)}}
+    runs-on: ubuntu-24.04
+    name: ${{ matrix.test-path.name }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Python & Poetry Environment
+        uses: exasol/python-toolbox/.github/actions/[email protected]
+        with:
+          python-version: '${{ matrix.python_version }}'
+
+      - name: Allow unprivileged user namespaces
+        run: sudo sysctl -w kernel.apparmor_restrict_unprivileged_userns=0
+
+      - name: Run test ${{ matrix.test-path.name }}
+        run: poetry run -- python ${{ matrix.test-path.path }}
+        env: # Set the secret as an input
+          DOCKER_USER: ${{ secrets.DOCKER_USERNAME }}
+          DOCKER_PASSWD: ${{ secrets.DOCKER_PASSWORD }}

.github/workflows/slow-checks.yml

@@ -27,46 +27,15 @@ jobs:
       - name: Tests
         run: echo "Slow tests approved"
 
-  prep-testbed:
-    needs: slow-tests-approval
-    runs-on: ubuntu-24.04
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Setup Python & Poetry Environment
-        uses: exasol/python-toolbox/.github/actions/[email protected]
-
-      - id: set-matrix
-        run: |
-          ALL_TESTS=`poetry run -- nox -s integration-test-list`
-          echo "matrix=$ALL_TESTS" >> "$GITHUB_OUTPUT"
-    outputs:
-      matrix: ${{ steps.set-matrix.outputs.matrix }}
 
-  test-all:
-    needs: prep-testbed
-    strategy:
-      fail-fast: false
-      matrix:
-        python_version:
-          - "3.10"
-        test-path: ${{fromJson(needs.prep-testbed.outputs.matrix)}}
-    runs-on: ubuntu-24.04
-    name: ${{ matrix.test-path.name }}
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Setup Python & Poetry Environment
-        uses: exasol/python-toolbox/.github/actions/[email protected]
-        with:
-          python-version: '${{ matrix.python_version }}'
-
-      - name: Allow unprivileged user namespaces
-        run: sudo sysctl -w kernel.apparmor_restrict_unprivileged_userns=0
+  integration-tests:
+    needs: slow-tests-approval
+    name: Integrations
+    uses: ./.github/workflows/integration-tests.yml
+    secrets: inherit
 
-      - name: Run test ${{ matrix.test-path.name }}
-        run: poetry run -- python ${{ matrix.test-path.path }}
-        env: # Set the secret as an input
-          DOCKER_USER: ${{ secrets.DOCKER_USERNAME }}
-          DOCKER_PASSWD: ${{ secrets.DOCKER_PASSWORD }}
+  gpu-tests:
+    needs: slow-tests-approval
+    name: Integrations
+    uses: ./.github/workflows/gpu-tests.yml
+    secrets: inherit

doc/changes/unreleased.md

@@ -6,3 +6,7 @@
 
 ## Refactorings
  - #252: Update Github workflows from Exasol toolbox
+
+## Features
+
+ - #283: Added GPU Support for `run-db-test`

doc/user_guide/user_guide.md

@@ -154,6 +154,13 @@ With this additional option, `exaslct` won't build and export the container agai
 1. `exaslct` won't check if the given container file is compatible with the given flavor path. If this is not the case, the tests probably will fail.
 2. As `exaslct` cannot check if the file was changed during different executions of `run-db-test` with options `--reuse-uploaded-container` or `--reuse-test-environment`, `exaslct` will always skip the upload if those options are set and the file already exists in BucketFS.
 
+### Testing with an accelerator
+
+If your test environment fulfills the requirements, you can run tests with an accelerator. Currently, only `nvidia` is supported:
+```bash
+exaslct run-db-test --flavor-path=flavors/<flavor-name> --accelerator nvidia
+```
+This will launch the underlying docker-db with the necessary options to enable the `NVIDIA` accelerator.
 
 ## Cleaning up after you are finished
 

exasol/slc/api/run_db_tests.py

@@ -35,6 +35,7 @@
 from exasol.slc.internal.tasks.test.test_container_content import (
     build_test_container_content,
 )
+from exasol.slc.models.accelerator import Accelerator, defaultAccelerator
 from exasol.slc.models.compression_strategy import (
     CompressionStrategy,
     defaultCompressionStrategy,
@@ -103,6 +104,7 @@ def run_db_test(
     log_level: Optional[str] = None,
     use_job_specific_log_file: bool = True,
     compression_strategy: CompressionStrategy = defaultCompressionStrategy(),
+    accelerator: Accelerator = defaultAccelerator(),
 ) -> AllTestsResult:
     """
     This command runs the integration tests in local docker-db.
@@ -155,6 +157,12 @@ def run_db_test(
         if external_exasol_ssh_port is None:
             raise api_errors.MissingArgumentError("external_exasol_ssh_port")
 
+    docker_runtime = None
+    if accelerator == Accelerator.NVIDA:
+        additional_db_parameter += ("-enableAcceleratorDeviceDetection=1",)
+        docker_runtime = "nvidia"
+        docker_environment_variable += ("NVIDIA_VISIBLE_DEVICES=all",)
+
     def root_task_generator() -> DependencyLoggerBaseTask:
         return generate_root_task(
             task_class=TestContainer,
@@ -199,6 +207,7 @@ def root_task_generator() -> DependencyLoggerBaseTask:
             additional_db_parameter=additional_db_parameter,
             test_container_content=build_test_container_content(test_container_folder),
             compression_strategy=compression_strategy,
+            docker_runtime=docker_runtime,
             docker_environment_variables=docker_environment_variable,
         )
 

exasol/slc/models/accelerator.py

@@ -0,0 +1,19 @@
+from enum import Enum
+from typing import List
+
+
+class Accelerator(Enum):
+    """
+    This enum serves as a definition of values for possible accelerators for `run-db-tests`.
+    """
+
+    NONE = "none"
+    NVIDA = "nvidia"
+
+
+def defaultAccelerator() -> Accelerator:
+    return Accelerator.NONE
+
+
+def acceleratorValues() -> List[str]:
+    return [a.value for a in Accelerator]

exasol/slc/tool/commands/run_db_tests.py

@@ -25,6 +25,11 @@
 
 from exasol.slc import api
 from exasol.slc.api import api_errors
+from exasol.slc.models.accelerator import (
+    Accelerator,
+    acceleratorValues,
+    defaultAccelerator,
+)
 from exasol.slc.models.compression_strategy import CompressionStrategy
 from exasol.slc.tool.cli import cli
 from exasol.slc.tool.options.export_options import export_options
@@ -149,6 +154,13 @@
 @add_options(system_options)
 @add_options(luigi_logging_options)
 @add_options(export_options)
+@click.option(
+    "--accelerator",
+    type=click.Choice(acceleratorValues()),
+    show_default=True,
+    default=defaultAccelerator().value,
+    help=f"""Accelerator to be enabled for tests in docker-db. Possible values: {acceleratorValues()}""",
+)
 def run_db_test(
     flavor_path: Tuple[str, ...],
     release_goal: Tuple[str, ...],
@@ -209,6 +221,7 @@ def run_db_test(
     log_level: Optional[str],
     use_job_specific_log_file: bool,
     compression_strategy: str,
+    accelerator: str,
 ):
     """
     This command runs the integration tests in local docker-db.
@@ -279,6 +292,7 @@ def run_db_test(
                 use_job_specific_log_file=use_job_specific_log_file,
                 compression_strategy=CompressionStrategy[compression_strategy.upper()],
                 docker_environment_variable=docker_environment_variable,
+                accelerator=Accelerator[accelerator.upper()],
             )
             if result.command_line_output_path.exists():
                 with result.command_line_output_path.open("r") as f:

noxfile.py

@@ -1,4 +1,6 @@
 import json
+from argparse import ArgumentParser
+from enum import Enum
 
 import nox
 
@@ -11,11 +13,38 @@
 from noxconfig import PROJECT_CONFIG
 
 
+class TestSet(Enum):
+    GPU_ONLY = "gpu-only"
+    DEFAULT = "default"
+
+
 @nox.session(name="integration-test-list", python=False)
 def run_integration_test_list(session: nox.Session):
     """
     Returns the test files under directory test (without walking subdirectories) as JSON list.
     """
+    test_set_values = [ts.value for ts in TestSet]
+    parser = ArgumentParser(
+        usage=f"nox -s {session.name} -- --test-set {{{','.join(test_set_values)}}}"
+    )
+    parser.add_argument(
+        "--test-set",
+        type=TestSet,
+        required=True,
+        help="Test set name",
+    )
+    args = parser.parse_args(session.posargs)
     test_path = PROJECT_CONFIG.root / "test"
-    tests = [{"path": str(t), "name": t.stem} for t in test_path.glob("test_*.py")]
+    if args.test_set == TestSet.GPU_ONLY:
+        tests = [
+            {"path": str(t), "name": t.stem}
+            for t in test_path.glob("test_*.py")
+            if "gpu" in t.name
+        ]
+    else:
+        tests = [
+            {"path": str(t), "name": t.stem}
+            for t in test_path.glob("test_*.py")
+            if "gpu" not in t.name
+        ]
     print(json.dumps(tests))