Fix zombie process exception (#133)

* Fix zombie process exception

* Fix flake8 checks

* Retrigger tests

* Pin psutil version to 5.9.5

* Unpin psutil version and add sleep in local integration tests

Author: sachanub

src/sagemaker_inference/model_server.py

@@ -273,6 +273,8 @@ def _retrieve_mms_server_process():
     mms_server_processes = list()
 
     for process in psutil.process_iter():
+        if process.status() == psutil.STATUS_ZOMBIE:
+            continue
         if MMS_NAMESPACE in process.cmdline():
             mms_server_processes.append(process)
 

test/integration/local/test_dummy_multi_model.py

@@ -46,6 +46,7 @@ def container():
             except:  # noqa: E722
                 attempts += 1
                 pass
+        time.sleep(60)
         yield proc.pid
     finally:
         subprocess.check_call("docker rm -f sagemaker-inference-toolkit-test".split())

test/integration/local/test_mxnet_multi_model.py

@@ -46,6 +46,7 @@ def container():
             except:  # noqa: E722
                 attempts += 1
                 pass
+        time.sleep(60)
         yield proc.pid
     finally:
         subprocess.check_call("docker rm -f sagemaker-inference-toolkit-test".split())