[RAPTOR-15086] Add vLLM Dockerfile build tests

tests/functional/test_vllm/README.md

@@ -0,0 +1,156 @@
+# vLLM Dockerfile Tests
+
+Validates that the vLLM Docker environment builds correctly and contains all required components.
+
+## Overview
+
+This test suite ensures the vLLM Dockerfile:
+- ✅ Builds successfully without errors
+- ✅ Contains required components (Python 3.12, vLLM dependencies)
+- ✅ Runs with proper security (non-root user)
+
+## Test Structure
+
+```
+test_vllm/
+├── test_dockerfile.py           # 3 tests
+└── README.md                    # This file
+```
+
+### Tests
+
+**TestVLLMDockerBuild** - 3 comprehensive tests (requires Docker):
+
+1. `test_image_builds_successfully` - Validates Docker build completes
+2. `test_image_has_required_components` - Verifies Python 3.12, virtualenv, paths
+3. `test_container_runs_as_non_root` - Security validation (uid 1000)
+
+## Running Tests
+
+### As Part of All Functional Tests (Recommended):
+```bash
+make functional-tests  # Runs all functional tests including vLLM
+```
+
+### Directly with pytest:
+```bash
+# Run only vLLM tests
+pytest tests/functional/test_vllm/test_dockerfile.py -v
+
+# Or as part of all functional tests
+pytest tests/functional/ -v
+```
+
+## CI/CD Integration
+
+### Jenkins
+```groovy
+stage('Test vLLM') {
+    steps {
+        sh 'pytest tests/functional/test_vllm/test_dockerfile.py -v'
+    }
+}
+```
+
+### GitHub Actions
+```yaml
+- name: Test vLLM Dockerfile
+  run: pytest tests/functional/test_vllm/test_dockerfile.py -v
+```
+
+### GitLab CI
+```yaml
+test:vllm:
+  script:
+    - pytest tests/functional/test_vllm/test_dockerfile.py -v
+```
+
+## Performance
+
+| Run Type | Time |
+|----------|------|
+| First run (with image download) | 5-15 min |
+| Cached run | 3-5 min |
+
+The Docker image is built once and reused across all tests in the class.
+
+## Requirements
+
+### On Your Host Machine (to run tests):
+- Docker daemon running
+- Python 3.8+ (for pytest test runner)
+- pytest, docker Python packages
+
+### Inside the Docker Container (what gets tested):
+- Python 3.12 (installed by Dockerfile)
+- vLLM dependencies (from base image)
+- ~15GB free disk space for image
+
+## Troubleshooting
+
+### Docker not running:
+```bash
+# Error: "Docker is not available or accessible"
+# Solution:
+# macOS/Windows: Start Docker Desktop
+# Linux: sudo systemctl start docker
+```
+
+### Insufficient disk space:
+```bash
+# Error: "no space left on device"
+# Solution:
+docker system prune -a  # Remove unused images
+docker images           # Check current images
+df -h                   # Check disk space
+```
+
+### Build fails - Base image unavailable:
+```bash
+# Error: "pull access denied" or "manifest unknown"
+# Context: Image pulled from Docker Hub: vllm/vllm-openai:v0.11.0
+# Solution:
+# 1. Check internet connection
+# 2. Verify base image exists: docker pull vllm/vllm-openai:v0.11.0
+# 3. If behind corporate proxy, configure Docker proxy settings
+# 4. If authentication required, login: docker login
+```
+
+### Build fails - Authentication required:
+```bash
+# Error: "pull access denied for vllm/vllm-openai" with "authentication required"
+# Solution:
+# 1. Login to Docker Hub: docker login
+# 2. Or use credentials: docker login -u <username> -p <password>
+# 3. If using private registry, configure credentials in Docker config
+```
+
+### Build fails - Dependency installation:
+```bash
+# Error: "E: Unable to locate package python3.12"
+# Context: Dockerfile specifies Python 3.12 installation (line 4)
+# Solution:
+# 1. Verify base image hasn't changed: docker pull vllm/vllm-openai:v0.11.0
+# 2. Check if Dockerfile was modified in public_dropin_gpu_environments/vllm/
+# 3. Try rebuilding without cache: docker build --no-cache
+```
+
+### Container fails to start:
+```bash
+# Error: "Container did not start in time"
+# Solution:
+# 1. Check Docker daemon logs: docker logs <container-id>
+# 2. Verify sufficient resources (CPU/memory)
+# 3. Check for port conflicts
+```
+
+
+## Design Philosophy
+
+These tests follow best practices:
+- **No redundancy**: Each test validates something unique
+- **Fast feedback**: Docker build happens once, reused across tests
+- **CI-friendly**: Integrates seamlessly with existing test infrastructure
+- **Clear failures**: Detailed error messages for debugging
+
+

tests/functional/test_vllm/test_dockerfile.py

@@ -0,0 +1,111 @@
+"""
+Copyright 2025 DataRobot, Inc. and its affiliates.
+All rights reserved.
+This is proprietary source code of DataRobot, Inc. and its affiliates.
+Released under the terms of DataRobot Tool and Utility Agreement.
+
+vLLM Dockerfile build and validation tests.
+
+These tests validate that the vLLM Docker image can be built successfully
+and contains all required components for running in production.
+"""
+
+import os
+import time
+
+import docker
+import pytest
+
+from tests.constants import PUBLIC_DROPIN_GPU_ENVS_PATH
+
+
+class TestVLLMDockerBuild:
+    """Docker build tests for vLLM environment (requires Docker)."""
+
+    VLLM_PATH = os.path.join(PUBLIC_DROPIN_GPU_ENVS_PATH, "vllm")
+    IMAGE_TAG = "vllm-test:latest"
+
+    @pytest.fixture(scope="class")
+    def docker_client(self):
+        """Provides a Docker client, failing if Docker is not available."""
+        try:
+            client = docker.from_env()
+            client.ping()
+            return client
+        except Exception as e:
+            pytest.fail(f"Docker is not available or accessible: {e}")
+
+    @pytest.fixture(scope="class")
+    def vllm_image(self, docker_client):
+        """Builds the vLLM Docker image once per test class."""
+        try:
+            image, build_logs = docker_client.images.build(
+                path=self.VLLM_PATH,
+                tag=self.IMAGE_TAG,
+                rm=True,
+                forcerm=True,
+            )
+        except docker.errors.BuildError as e:
+            build_log = "".join([log.get("stream", "") for log in e.build_log or []])
+            pytest.fail(
+                f"Docker build failed: {e}\nBuild Log (last 50 lines):\n{build_log[-5000:]}"
+            )
+
+        yield image
+
+        # Cleanup: remove the image after tests
+        try:
+            docker_client.images.remove(image.id, force=True)
+        except docker.errors.APIError:
+            pass  # Ignore errors during cleanup
+
+    @pytest.fixture(scope="class")
+    def vllm_container(self, docker_client, vllm_image):
+        """Creates and runs a container that is shared across tests in this class."""
+        container = docker_client.containers.run(
+            vllm_image.id,
+            command="sleep 600",  # Keep container running for tests
+            detach=True,
+        )
+        # Wait for the container to be in the 'running' state
+        for _ in range(10):
+            container.reload()
+            if container.status == "running":
+                break
+            time.sleep(1)
+        else:
+            pytest.fail("Container did not start in time")
+
+        yield container
+        # Reliable cleanup
+        try:
+            container.remove(force=True)
+        except docker.errors.APIError:
+            pass  # Ignore errors during cleanup
+
+    def test_image_builds_successfully(self, vllm_image):
+        """Verify the Docker image was built and has an ID."""
+        assert vllm_image is not None
+        assert vllm_image.id is not None
+
+    def test_image_has_required_components(self, vllm_container):
+        """Verify the built image contains required paths and packages."""
+        required_paths = ["/opt/code", "/opt/venv", "/opt/.home"]
+        for path in required_paths:
+            exit_code, _ = vllm_container.exec_run(f"test -e {path}")
+            assert exit_code == 0, f"Required path not found in container: {path}"
+
+        # Check for Python 3.12
+        exit_code, _ = vllm_container.exec_run("python3.12 --version")
+        assert exit_code == 0, "Python 3.12 not found in container"
+
+        # Check that virtualenv was created
+        exit_code, _ = vllm_container.exec_run("/opt/venv/bin/python --version")
+        assert exit_code == 0, "Virtualenv Python not found in container"
+
+    def test_container_runs_as_non_root(self, vllm_container):
+        """Verify the container runs with the correct non-root user."""
+        exit_code, output = vllm_container.exec_run("whoami")
+        assert exit_code == 0
+        # The user is 'datarobot' in the Dockerfile
+        assert output.decode().strip() == "datarobot"