[Cherry Pick] ci: Fix Windows CI Errors

qa/L0_backend_python/lifecycle/lifecycle_test.py

@@ -68,7 +68,7 @@ def setUp(self):
         self._shm_leak_detector = shm_util.ShmLeakDetector()
 
     def _get_metrics(self):
-        metrics_url = "http://localhost:8002/metrics"
+        metrics_url = f"http://{_tritonserver_ipaddr}:8002/metrics"
         r = requests.get(metrics_url)
         r.raise_for_status()
         return r.text
@@ -305,7 +305,9 @@ def test_triton_grpc_error_cancel(self):
         shape = [2, 2]
         number_of_requests = 1
         user_data = UserData()
-        triton_server_url = "localhost:8001"  # Replace with your Triton server address
+        triton_server_url = (
+            f"{_tritonserver_ipaddr}:8001"  # Replace with your Triton server address
+        )
         stream_end = False
         triton_client = grpcclient.InferenceServerClient(triton_server_url)
 

qa/L0_backend_python/response_sender/response_sender_complete_final_test.py

@@ -31,6 +31,10 @@
 import numpy as np
 import tritonclient.grpc as grpcclient
 
+# By default, find tritonserver on "localhost", but for windows tests
+# we overwrite the IP address with the TRITONSERVER_IPADDR envvar
+_tritonserver_ipaddr = os.environ.get("TRITONSERVER_IPADDR", "localhost")
+
 
 class ResponseSenderTest(unittest.TestCase):
     def _generate_streaming_callback_and_responses_pair(self):
@@ -53,7 +57,7 @@ def test_respond_after_complete_final(self):
         inputs[0].set_data_from_numpy(input0_np)
 
         callback, responses = self._generate_streaming_callback_and_responses_pair()
-        with grpcclient.InferenceServerClient("localhost:8001") as client:
+        with grpcclient.InferenceServerClient(f"{_tritonserver_ipaddr}:8001") as client:
             client.start_stream(callback)
             client.async_stream_infer(model_name, inputs)
             client.stop_stream()

qa/L0_backend_python/response_sender/response_sender_test.py

@@ -24,12 +24,17 @@
 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
+import os
 import unittest
 
 import numpy as np
 import tritonclient.grpc as grpcclient
 from tritonclient.utils import InferenceServerException
 
+# By default, find tritonserver on "localhost", but for windows tests
+# we overwrite the IP address with the TRITONSERVER_IPADDR envvar
+_tritonserver_ipaddr = os.environ.get("TRITONSERVER_IPADDR", "localhost")
+
 
 class ResponseSenderTest(unittest.TestCase):
     _inputs_parameters_zero_response_pre_return = {
@@ -148,7 +153,7 @@ def callback(result, error):
 
     def _infer_parallel(self, model_name, parallel_inputs):
         callback, responses = self._generate_streaming_callback_and_responses_pair()
-        with grpcclient.InferenceServerClient("localhost:8001") as client:
+        with grpcclient.InferenceServerClient(f"{_tritonserver_ipaddr}:8001") as client:
             client.start_stream(callback)
             for inputs in parallel_inputs:
                 client.async_stream_infer(model_name, inputs)

qa/L0_backend_python/response_sender/test.sh

@@ -94,8 +94,7 @@ if [ $? -ne 0 ]; then
 fi
 set -e
 
-kill $SERVER_PID
-wait $SERVER_PID
+kill_server
 
 #
 # Test response sender to raise exception on response after complete final flag
@@ -125,8 +124,7 @@ if [ $? -ne 0 ]; then
 fi
 set -e
 
-kill $SERVER_PID
-wait $SERVER_PID
+kill_server
 
 #
 # Test async response sender under decoupled / non-decoupled

qa/L0_batcher/queue_timeout_test.py

@@ -27,18 +27,23 @@
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 import concurrent.futures
+import os
 import time
 import unittest
 
 import numpy as np
 import tritonclient.grpc as grpcclient
 from tritonclient.utils import InferenceServerException
 
+# By default, find tritonserver on "localhost", but for windows tests
+# we overwrite the IP address with the TRITONSERVER_IPADDR envvar
+_tritonserver_ipaddr = os.environ.get("TRITONSERVER_IPADDR", "localhost")
+
 
 class TestMaxQueueDelayTimeout(unittest.TestCase):
     def setUp(self):
         # Initialize client
-        self._triton = grpcclient.InferenceServerClient("localhost:8001")
+        self._triton = grpcclient.InferenceServerClient(f"{_tritonserver_ipaddr}:8001")
 
     def _get_inputs(self, batch_size):
         self.assertIsInstance(batch_size, int)

qa/L0_batcher/test.sh

@@ -85,6 +85,10 @@ if [[ -v WSL_DISTRO_NAME ]] || [[ -v MSYSTEM ]]; then
     BACKEND_DIR=${BACKEND_DIR:=C:/tritonserver/backends}
     SERVER=${SERVER:=/mnt/c/tritonserver/bin/tritonserver.exe}
     export WSLENV=$WSLENV:TRITONSERVER_DELAY_SCHEDULER
+    TEST_WINDOWS=1
+    # DLIS-7683 This test fails performance-related response time parameters
+    # when using HTTP protocol. Use gRPC protocol for now as a WAR.
+    export USE_GRPC=1
 else
     MODELDIR=${MODELDIR:=`pwd`}
     DATADIR=${DATADIR:="/data/inferenceserver/${REPO_VERSION}"}
@@ -601,7 +605,7 @@ done
 TEST_CASE=test_multi_batch_preserve_ordering
 
 # Skip test for Windows. Trace file concats at 8192 chars on Windows.
-if  [[ ! -v WSL_DISTRO_NAME ]] || [[ ! -v MSYSTEM ]]; then
+if  [ $TEST_WINDOWS -eq 0 ]; then
     rm -fr ./custom_models && mkdir ./custom_models && \
         cp -r ../custom_models/custom_zero_1_float32 ./custom_models/. && \
         mkdir -p ./custom_models/custom_zero_1_float32/1
@@ -754,7 +758,7 @@ mkdir -p models/dynamic_batch/1 && (cd models/dynamic_batch && \
 TEST_LOG="queue_timeout_test.log"
 SERVER_LOG="./queue_timeout_test.server.log"
 
-SERVER_ARGS="--model-repository=`pwd`/models --log-verbose=2"
+SERVER_ARGS="--model-repository=$MODELDIR/models --log-verbose=2 --backend-directory=${BACKEND_DIR}"
 run_server
 if [ "$SERVER_PID" == "0" ]; then
     echo -e "\n***\n*** Failed to start $SERVER\n***"
@@ -771,8 +775,7 @@ if [ $? -ne 0 ]; then
 fi
 set -e
 
-kill $SERVER_PID
-wait $SERVER_PID
+kill_server
 
 if [ $RET -eq 0 ]; then
     echo -e "\n***\n*** Test Passed\n***"

qa/L0_trt_compat/test.sh

@@ -34,7 +34,7 @@ if [ -z "$REPO_VERSION" ]; then
     echo -e "\n***\n*** Test Failed\n***"
     exit 1
 fi
-if [ ! -z "$TEST_REPO_ARCH" ]; then 
+if [ ! -z "$TEST_REPO_ARCH" ]; then
     REPO_VERSION=${REPO_VERSION}_${TEST_REPO_ARCH}
 fi
 

qa/L0_trt_plugin/test.sh

@@ -53,6 +53,7 @@ if [[ -v WSL_DISTRO_NAME ]] || [[ -v MSYSTEM ]]; then
     CUSTOMPLUGIN=${CUSTOMPLUGIN:=$MODELDIR/HardmaxPlugin.dll}
     BACKEND_DIR=${BACKEND_DIR:=C:/tritonserver/backends}
     SERVER=${SERVER:=/mnt/c/tritonserver/bin/tritonserver.exe}
+    TEST_WINDOWS=1
 else
     DATADIR=${DATADIR:="/data/inferenceserver/${REPO_VERSION}"}
     MODELDIR=${MODELDIR:=`pwd`/models}
@@ -135,7 +136,8 @@ SERVER_LD_PRELOAD=$CUSTOMPLUGIN
 SERVER_ARGS=$SERVER_ARGS_BASE
 SERVER_LOG="./inference_server_$LOG_IDX.log"
 
-if  [[ ! -v WSL_DISTRO_NAME ]] || [[ ! -v MSYSTEM ]]; then
+# Skip test for Windows
+if  [ $TEST_WINDOWS -eq 0 ]; then
     run_server
     if [ "$SERVER_PID" == "0" ]; then
         echo -e "\n***\n*** Failed to start $SERVER\n***"

qa/common/infer_util.py

@@ -701,7 +701,7 @@ def inferAndCheckResults(
                                 )
                                 if cidx == expected0_sort_idx[b][idx]:
                                     tester.assertEqual(
-                                        ctuple[2],
+                                        ctuple[2].strip("\r"),
                                         "label{}".format(expected0_sort_idx[b][idx]),
                                     )
                             elif result_name == OUTPUT1: