Feedback: Ruth

Author: israel-hdez

tests/model_serving/model_server/inference_graph/conftest.py

@@ -42,7 +42,6 @@ def dog_breed_inference_graph(
         nodes=nodes,
     ) as inference_graph:
         inference_graph.wait_for_condition(condition=inference_graph.Condition.READY, status="True")
-        time.sleep(5)  # This wait helps OpenShift to fully enable the Route, if the InferenceGraph is exposed
         yield inference_graph
 
 

utilities/inference_utils.py

@@ -404,7 +404,7 @@ def run_inference_flow(
         except JSONDecodeError:
             return {"output": out}
 
-    @retry(wait_timeout=Timeout.TIMEOUT_30SEC, sleep=5)
+    @retry(wait_timeout=Timeout.TIMEOUT_30SEC, sleep=5, exceptions_dict={AssertionError:[]})
     def run_inference(self, cmd: str) -> str:
         """
         Run inference command
@@ -439,6 +439,9 @@ def run_inference(self, cmd: str) -> str:
 
         else:
             res, out, err = run_command(command=shlex.split(cmd), verify_stderr=False, check=False)
+            if res and "http/1.0 503 service unavailable" in out.lower():
+                LOGGER.info(f"The Route for {self.get_inference_url()} is not ready yet (503 error)")
+                raise AssertionError("The Route is not ready yet")
 
         if not res:
             raise ValueError(f"Inference failed with error: {err}\nOutput: {out}\nCommand: {cmd}")