allow rerunning test without tearing down kind cluster

tests/system-test-kind.sh

@@ -1,11 +1,16 @@
 #!/usr/bin/env bash
 
-set -e
+set -xe
 
-kind create cluster --name=substratus-test
-# trap "kind delete cluster --name=substratus-test" EXIT
+if kind get clusters | grep -q substratus-test; then
+  echo "Cluster substratus-tests already exists.. reusing it"
+  else
+  kind create cluster --name substratus-test
+fi
 
-skaffold run
+if ! kubectl get deployment proxy-controller; then
+  skaffold run
+fi
 
 kubectl wait --for=condition=available --timeout=30s deployment/proxy-controller
 
@@ -30,7 +35,9 @@ fi
 SCRIPT_DIR=$(dirname "$0")
 VENV_DIR=$SCRIPT_DIR/.venv
 
-python3 -m venv "$VENV_DIR"
+if [ ! -d "$VENV_DIR" ]; then
+  python3 -m venv "$VENV_DIR"
+fi
 source "$VENV_DIR/bin/activate"
 pip3 install openai==1.2.3
 
@@ -51,8 +58,7 @@ requests=500
 echo "Send $requests requests in parallel to stapi backend using openai python client and threading"
 python3 $SCRIPT_DIR/test_openai_embedding.py \
   --requests $requests \
-  --model text-embedding-ada-002 \
-  --client-per-thread False
+  --model text-embedding-ada-002
 
 replicas=$(kubectl get deployment stapi-minilm-l6-v2 -o jsonpath='{.spec.replicas}')
 if [ "$replicas" -ge 2 ]; then

tests/test_openai_embedding.py

@@ -7,7 +7,7 @@
 parser.add_argument("--requests", type=int, default=60)
 parser.add_argument("--model", type=str, default="text-embedding-ada-002")
 parser.add_argument("--text", type=str, default="Generate an embedding for me")
-parser.add_argument("--client-per-thread", type=bool, default=False)
+# parser.add_argument("--client-per-thread", type=bool, default=False)
 args = parser.parse_args()
 
 def create_client():
@@ -16,17 +16,17 @@ def create_client():
         base_url=args.base_url,
     )
 
-client = create_client()
+
+# client = create_client()
 
 def embedding_request(index: int):
     print (f"Request {index} of {args.requests}")
-    if args.client_per_thread:
-        client = create_client()
+    client = create_client()
     embedding = client.embeddings.create(model=args.model, input=args.text)
     print (f"Finished {index} of {args.requests}")
     return embedding
 
 with concurrent.futures.ThreadPoolExecutor(max_workers=args.requests) as executor:
     futures = [executor.submit(embedding_request, i+1) for i in range(args.requests)]
-    results = [future.result() for future in concurrent.futures.as_completed(futures, timeout=600)]
+    results = [future.result() for future in concurrent.futures.as_completed(futures, timeout=900)]
     assert len(results) == args.requests