fix: simplify GMS layout state and harden GPU-backed flows (#7006)

Signed-off-by: mohammedabdulwahhab <furkhan324@berkeley.edu> Signed-off-by: Hannah Zhang <hannahz@nvidia.com> Co-authored-by: mohammedabdulwahhab <furkhan324@berkeley.edu> Co-authored-by: hhzhang16 <54051230+hhzhang16@users.noreply.github.com>

fix: simplify GMS layout state and harden GPU-backed flows (#7006)
Signed-off-by: mohammedabdulwahhab <furkhan324@berkeley.edu> Signed-off-by: Hannah Zhang <hannahz@nvidia.com> Co-authored-by: mohammedabdulwahhab <furkhan324@berkeley.edu> Co-authored-by: hhzhang16 <54051230+hhzhang16@users.noreply.github.com>
91a8c6af · Schwinn Saereesitthipitak · GitHub · dd7ceb4a · 91a8c6af · 91a8c6af
Unverified Commit 91a8c6af authored Apr 01, 2026 by Schwinn Saereesitthipitak Committed by GitHub Apr 02, 2026
20 changed files
--- a/tests/fault_tolerance/gpu_memory_service/test_failover_lock.py
+++ b/tests/fault_tolerance/gpu_memory_service/test_failover_lock.py
@@ -20,7 +20,6 @@ pytestmark = [
    pytest.mark.pre_merge,
    pytest.mark.unit,
    pytest.mark.gpu_0,
-    pytest.mark.fault_tolerance,
 ]



--- a/tests/gms/common/test_gms_client_memory_manager.py
+++ b/tests/gms/common/test_gms_client_memory_manager.py
--- a/tests/gms/common/test_gms_client_session.py
+++ b/tests/gms/common/test_gms_client_session.py
--- a/tests/gms/common/test_gms_client_transport.py
+++ b/tests/gms/common/test_gms_client_transport.py
--- a/tests/gms/common/test_gms_harness.py
+++ b/tests/gms/common/test_gms_harness.py
--- a/tests/gms/common/test_gms_runtime_flows.py
+++ b/tests/gms/common/test_gms_runtime_flows.py
--- a/tests/gms/common/test_gms_server_transport_failures.py
+++ b/tests/gms/common/test_gms_server_transport_failures.py
--- a/tests/gms/common/test_gms_sglang_patches.py
+++ b/tests/gms/common/test_gms_sglang_patches.py
--- a/tests/gms/common/test_gms_torch_allocator.py
+++ b/tests/gms/common/test_gms_torch_allocator.py
--- a/tests/fault_tolerance/gpu_memory_service/conftest.py
+++ b/tests/fault_tolerance/gpu_memory_service/conftest.py
--- a/tests/fault_tolerance/gpu_memory_service/utils/__init__.py
+++ b/tests/fault_tolerance/gpu_memory_service/utils/__init__.py
--- a/tests/gms/harness/external_weight_writer.py
+++ b/tests/gms/harness/external_weight_writer.py
--- a/tests/gms/harness/gms.py
+++ b/tests/gms/harness/gms.py
--- a/tests/gms/harness/runtime.py
+++ b/tests/gms/harness/runtime.py
--- a/tests/fault_tolerance/gpu_memory_service/utils/sglang.py
+++ b/tests/fault_tolerance/gpu_memory_service/utils/sglang.py
--- a/tests/fault_tolerance/gpu_memory_service/utils/vllm.py
+++ b/tests/fault_tolerance/gpu_memory_service/utils/vllm.py
--- a/tests/gms/integration/__init__.py
+++ b/tests/gms/integration/__init__.py
--- a/tests/gms/integration/test_external_weight_mgr.py
+++ b/tests/gms/integration/test_external_weight_mgr.py
--- a/tests/gms/integration/test_gms_sglang_memory_saver.py
+++ b/tests/gms/integration/test_gms_sglang_memory_saver.py
--- a/tests/gms/integration/test_gms_shadow_failover.py
+++ b/tests/gms/integration/test_gms_shadow_failover.py