Refactor Ray Serve app startup in the tests

to use fixed port and route prefix

aana/tests/deployments/test_hf_blip2_deployment.py

@@ -1,5 +1,4 @@
 from importlib import resources
-import random
 import pytest
 import ray
 from ray import serve
@@ -13,9 +12,10 @@ def ray_setup(deployment):
     # Setup ray environment and serve
     ray.init(ignore_reinit_error=True)
     app = deployment.bind()
-    # random port from 30000 to 40000
-    port = random.randint(30000, 40000)
-    handle = serve.run(app, port=port)
+    port = 34422
+    test_name = deployment.name
+    route_prefix = f"/{test_name}"
+    handle = serve.run(app, port=port, name=test_name, route_prefix=route_prefix)
     return handle
 
 

aana/tests/deployments/test_vllm_deployment.py

@@ -1,4 +1,3 @@
-import random
 import pytest
 import ray
 from ray import serve
@@ -27,9 +26,10 @@ def ray_setup(deployment):
     # Setup ray environment and serve
     ray.init(ignore_reinit_error=True)
     app = deployment.bind()
-    # random port from 30000 to 40000
-    port = random.randint(30000, 40000)
-    handle = serve.run(app, port=port)
+    port = 34422
+    test_name = deployment.name
+    route_prefix = f"/{test_name}"
+    handle = serve.run(app, port=port, name=test_name, route_prefix=route_prefix)
     return handle
 
 

aana/tests/test_app.py

@@ -1,5 +1,4 @@
 import json
-import random
 import pytest
 import ray
 from ray import serve
@@ -69,7 +68,7 @@ async def lower(self, text: str) -> dict:
 
 
 @pytest.fixture(scope="session")
-def ray_setup():
+def ray_setup(request):
     """
     Setup the Ray environment and serve the endpoints.
 
@@ -79,27 +78,29 @@ def ray_setup():
     """
     ray.init(ignore_reinit_error=True)
     server = RequestHandler.bind(endpoints, nodes, context)
-    # random port from 30000 to 40000
-    port = random.randint(30000, 40000)
-    handle = serve.run(server, port=port)
-    return handle, port
+    port = 34422
+    test_name = request.node.name
+    route_prefix = f"/{test_name}"
+    handle = serve.run(server, port=port, name=test_name, route_prefix=route_prefix)
+    return handle, port, route_prefix
 
 
 def test_app(ray_setup):
     """
     Test the Ray Serve app.
     """
-    handle, port = ray_setup
+    handle, port, route_prefix = ray_setup
 
     # Check that the server is ready
-    response = requests.get(f"http://localhost:{port}/api/ready")
+    response = requests.get(f"http://localhost:{port}{route_prefix}/api/ready")
     assert response.status_code == 200
     assert response.json() == {"ready": True}
 
     # Test lowercase endpoint
     data = {"text": ["Hello World!", "This is a test."]}
     response = requests.post(
-        f"http://localhost:{port}/lowercase", data={"body": json.dumps(data)}
+        f"http://localhost:{port}{route_prefix}/lowercase",
+        data={"body": json.dumps(data)},
     )
     assert response.status_code == 200
     lowercase_text = response.json().get("lowercase_text")

aana/tests/test_app_streaming.py

@@ -1,5 +1,6 @@
 import json
-import random
+from sys import prefix
+import time
 from typing import AsyncGenerator
 import pytest
 import ray
@@ -77,7 +78,7 @@ async def lower_stream(self, text: str) -> AsyncGenerator[dict, None]:
 
 
 @pytest.fixture(scope="session")
-def ray_setup():
+def ray_setup(request):
     """
     Setup the Ray environment and serve the endpoints.
 
@@ -87,29 +88,30 @@ def ray_setup():
     """
     ray.init(ignore_reinit_error=True)
     server = RequestHandler.bind(endpoints, nodes, context)
-    # random port from 30000 to 40000
-    port = random.randint(30000, 40000)
-    handle = serve.run(server, port=port)
-    return handle, port
+    port = 34422
+    test_name = request.node.name
+    route_prefix = f"/{test_name}"
+    handle = serve.run(server, port=port, name=test_name, route_prefix=route_prefix)
+    return handle, port, route_prefix
 
 
 def test_app_streaming(ray_setup):
     """
     Test the Ray Serve app with streaming enabled.
     """
 
-    handle, port = ray_setup
+    handle, port, route_prefix = ray_setup
 
     # Check that the server is ready
-    response = requests.get(f"http://localhost:{port}/api/ready")
+    response = requests.get(f"http://localhost:{port}{route_prefix}/api/ready")
     assert response.status_code == 200
     assert response.json() == {"ready": True}
 
     # Test lowercase endpoint
     text = "Hello World, this is a test."
     data = {"text": text}
     response = requests.post(
-        f"http://localhost:{port}/lowercase",
+        f"http://localhost:{port}{route_prefix}/lowercase",
         data={"body": json.dumps(data)},
         stream=True,
     )