feat: support running docker runtime stresstest in CI (#6100)

Co-authored-by: Boxuan Li <liboxuan@connect.hku.hk>

openhands/core/config/sandbox_config.py

@@ -35,6 +35,10 @@ class SandboxConfig:
         remote_runtime_resource_factor: Factor to scale the resource allocation for remote runtime.
             Must be one of [1, 2, 4, 8]. Will only be used if the runtime is remote.
         enable_gpu: Whether to enable GPU.
+        docker_runtime_kwargs: Additional keyword arguments to pass to the Docker runtime when running containers.
+            This should be a JSON string that will be parsed into a dictionary.
+            Example in config.toml:
+            docker_runtime_kwargs = '{"mem_limit": "4g", "cpu_quota": 100000}'
     """
 
     remote_runtime_api_url: str = 'http://localhost:8000'
@@ -61,6 +65,7 @@ class SandboxConfig:
     close_delay: int = 900
     remote_runtime_resource_factor: int = 1
     enable_gpu: bool = False
+    docker_runtime_kwargs: str | None = None
 
     def defaults_to_dict(self) -> dict:
         """Serialize fields to a dict for the frontend, including type hints, defaults, and whether it's optional."""

openhands/runtime/impl/docker/docker_runtime.py

@@ -267,13 +267,11 @@ class DockerRuntime(ActionExecutionClient):
                 environment=environment,
                 volumes=volumes,
                 device_requests=(
-                    [docker.types.DeviceRequest(
-                        capabilities=[['gpu']],
-                        count=-1
-                    )]
+                    [docker.types.DeviceRequest(capabilities=[['gpu']], count=-1)]
                     if self.config.sandbox.enable_gpu
                     else None
                 ),
+                **(self.config.sandbox.docker_runtime_kwargs or {}),
             )
             self.log('debug', f'Container started. Server url: {self.api_url}')
             self.send_status_message('STATUS$CONTAINER_STARTED')

tests/runtime/conftest.py

@@ -215,6 +215,7 @@ def _load_runtime(
     use_workspace: bool | None = None,
     force_rebuild_runtime: bool = False,
     runtime_startup_env_vars: dict[str, str] | None = None,
+    docker_runtime_kwargs: dict[str, str] | None = None,
 ) -> Runtime:
     sid = 'rt_' + str(random.randint(100000, 999999))
 
@@ -226,6 +227,7 @@ def _load_runtime(
     config.run_as_openhands = run_as_openhands
     config.sandbox.force_rebuild_runtime = force_rebuild_runtime
     config.sandbox.keep_runtime_alive = False
+    config.sandbox.docker_runtime_kwargs = docker_runtime_kwargs
     # Folder where all tests create their own folder
     global test_mount_path
     if use_workspace:

tests/runtime/test_stress_docker_runtime.py

@@ -1,18 +1,21 @@
 """Stress tests for the DockerRuntime, which connects to the ActionExecutor running in the sandbox."""
 
-import pytest
-from conftest import TEST_IN_CI, _close_test_runtime, _load_runtime
+from conftest import _close_test_runtime, _load_runtime
 
 from openhands.core.logger import openhands_logger as logger
 from openhands.events.action import CmdRunAction
 
 
-@pytest.mark.skipif(
-    TEST_IN_CI,
-    reason='This test should only be run locally, not in CI.',
-)
 def test_stress_docker_runtime(temp_dir, runtime_cls, repeat=1):
-    runtime = _load_runtime(temp_dir, runtime_cls)
+    runtime = _load_runtime(
+        temp_dir,
+        runtime_cls,
+        docker_runtime_kwargs={
+            'cpu_period': 100000,  # 100ms
+            'cpu_quota': 100000,  # Can use 100ms out of each 100ms period (1 CPU)
+            'mem_limit': '4G',  # 4 GB of memory
+        },
+    )
 
     action = CmdRunAction(
         command='sudo apt-get update && sudo apt-get install -y stress-ng'
@@ -23,11 +26,9 @@ def test_stress_docker_runtime(temp_dir, runtime_cls, repeat=1):
     assert obs.exit_code == 0
 
     for _ in range(repeat):
-        # run stress-ng stress tests for 5 minutes
-        # FIXME: this would make Docker daemon die, even though running this
-        # command on its own in the same container is fine
-        action = CmdRunAction(command='stress-ng --all 1 -t 5m')
-        action.timeout = 600
+        # run stress-ng stress tests for 1 minute
+        action = CmdRunAction(command='stress-ng --all 1 -t 1m')
+        action.timeout = 120
         logger.info(action, extra={'msg_type': 'ACTION'})
         obs = runtime.run_action(action)
         logger.info(obs, extra={'msg_type': 'OBSERVATION'})