feat(tasks): wire e2e:gpu to bootstrap cluster with GPU support (#547)

Pass CLUSTER_GPU=1 inline in e2e:python:gpu's depends so that the
cluster is bootstrapped with --gpu when GPU e2e tests are run.

Add --gpu flag handling to cluster-bootstrap.sh and default
OPENSHELL_E2E_GPU_IMAGE to an empty string so the server resolves
the default sandbox image when no override is provided.

Signed-off-by: Evan Lezar <elezar@nvidia.com>

Author: elezar

e2e/python/conftest.py

@@ -96,10 +96,10 @@ def _run(sandbox: Sandbox, code: str) -> tuple[int, str, str]:
 
 @pytest.fixture(scope="session")
 def gpu_sandbox_spec() -> datamodel_pb2.SandboxSpec:
-    image = os.environ.get(
-        "OPENSHELL_E2E_GPU_IMAGE",
-        "ghcr.io/nvidia/openshell-community/sandboxes/nvidia-gpu:latest",
-    )
+    # Empty string defers image resolution to the server, which substitutes
+    # the configured default sandbox image.  Set OPENSHELL_E2E_GPU_IMAGE to
+    # override (e.g. a locally-built or registry-mirrored image).
+    image = os.environ.get("OPENSHELL_E2E_GPU_IMAGE", "")
     return datamodel_pb2.SandboxSpec(
         gpu=True,
         template=datamodel_pb2.SandboxTemplate(image=image),

tasks/scripts/cluster-bootstrap.sh

@@ -243,6 +243,10 @@ fi
 
 DEPLOY_CMD=(openshell gateway start --name "${CLUSTER_NAME}" --port "${GATEWAY_PORT}")
 
+if [ "${CLUSTER_GPU:-0}" = "1" ]; then
+  DEPLOY_CMD+=(--gpu)
+fi
+
 if [ -n "${GATEWAY_HOST:-}" ]; then
   DEPLOY_CMD+=(--gateway-host "${GATEWAY_HOST}")
 

tasks/test.toml

@@ -43,6 +43,6 @@ run = "uv run pytest -o python_files='test_*.py' -m 'not gpu' -n ${E2E_PARALLEL:
 
 ["e2e:python:gpu"]
 description = "Run Python GPU e2e tests"
-depends = ["python:proto", "cluster"]
+depends = ["python:proto", "CLUSTER_GPU=1 cluster"]
 env = { UV_NO_SYNC = "1", PYTHONPATH = "python" }
 run = "uv run pytest -o python_files='test_*.py' -m gpu -n ${E2E_PARALLEL:-1} e2e/python"