server : use pytest-xdist for server tests

.github/workflows/server-sanitize.yml

@@ -103,7 +103,7 @@ jobs:
           source .venv/bin/activate
           cd tools/server/tests
           export ${{ matrix.extra_args }}
-          ./tests.sh
+          PYTEST_WORKERS=1 ./tests.sh
 
       - name: Slow tests
         id: server_integration_tests_slow
@@ -112,4 +112,4 @@ jobs:
           source .venv/bin/activate
           cd tools/server/tests
           export ${{ matrix.extra_args }}
-          SLOW_TESTS=1 ./tests.sh
+          PYTEST_WORKERS=1 SLOW_TESTS=1 ./tests.sh

tools/server/tests/conftest.py

@@ -1,7 +1,19 @@
+import os
 import pytest
+from filelock import FileLock
 from utils import *
 
 
+@pytest.fixture(scope="session", autouse=True)
+def configure_worker_port(request):
+    worker_id = getattr(request.config, "workerinput", {}).get("workerid", "master")
+    if worker_id != "master":
+        worker_num = int(worker_id[2:])
+        base = 8080 + worker_num * 10
+        os.environ["BASE_PORT"] = str(base)
+        os.environ["PORT"] = str(base)
+
+
 # ref: https://stackoverflow.com/questions/22627659/run-code-before-and-after-each-test-in-py-test
 @pytest.fixture(autouse=True)
 def stop_server_after_each_test():
@@ -16,6 +28,10 @@ def stop_server_after_each_test():
 
 
 @pytest.fixture(scope="session", autouse=True)
-def load_server_presets():
+def load_server_presets(configure_worker_port, tmp_path_factory):
     # this will be run once per test session, before any tests
-    ServerPreset.load_all()
+
+    # serialize model downloads across parallel workers.
+    root_tmp_dir = tmp_path_factory.getbasetemp().parent
+    with FileLock(str(root_tmp_dir / "load_all.lock")):
+        ServerPreset.load_all()

tools/server/tests/requirements.txt

@@ -1,5 +1,7 @@
 aiohttp~=3.9.3
 pytest~=8.3.3
+pytest-xdist~=3.6
+filelock~=3.16
 numpy~=1.26.4
 openai~=2.14.0
 prometheus-client~=0.20.0

tools/server/tests/tests.sh

@@ -6,13 +6,16 @@ cd $SCRIPT_DIR
 
 set -eu
 
+WORKERS="${PYTEST_WORKERS:-auto}"
+
 if [ $# -lt 1 ]
 then
     if [[ "${SLOW_TESTS:-0}" == 1 ]]; then
-        pytest --durations=30 -v -x
+        # --dist=loadfile means that all tests in the same file will be sent to the same worker.
+        pytest --durations=30 -v -x -n "${WORKERS}" --dist=loadfile
     else
-        pytest --durations=30 -v -x -m "not slow"
+        pytest --durations=30 -v -x -n "${WORKERS}" --dist=loadfile -m "not slow"
     fi
 else
-    pytest --durations=30 "$@"
+    pytest --durations=30 -n "${WORKERS}" --dist=loadfile "$@"
 fi

tools/server/tests/unit/test_compat_anthropic.py

@@ -21,7 +21,7 @@ def create_server():
     global server
     server = ServerPreset.tinyllama2()
     server.model_alias = "tinyllama-2-anthropic"
-    server.server_port = 8082
+    server.server_port = server_base_port() + 2
     server.n_slots = 1
     server.n_ctx = 8192
     server.n_batch = 2048
@@ -34,7 +34,7 @@ def vision_server():
     server = ServerPreset.tinygemma3()
     server.offline = False  # Allow downloading the model
     server.model_alias = "tinygemma3-anthropic"
-    server.server_port = 8083  # Different port to avoid conflicts
+    server.server_port = server_base_port() + 3  # Different port to avoid conflicts
     server.n_slots = 1
     return server
 
@@ -1015,7 +1015,7 @@ def test_anthropic_thinking_with_reasoning_model(stream):
     server.jinja = True
     server.n_ctx = 8192
     server.n_predict = 1024
-    server.server_port = 8084
+    server.server_port = server_base_port() + 4
     server.start(timeout_seconds=600)  # large model needs time to download
 
     if stream:

tools/server/tests/unit/test_mcp_servers.py

@@ -37,7 +37,7 @@ def _start_server_with_mcp(mcp_json: str, **kwargs) -> ServerProcess:
     srv = ServerPreset.router()
     srv.server_tools = "all"
     srv.no_ui = True
-    srv.server_port = 8085  # avoid conflict with load_all() which uses 8080
+    srv.server_port = server_base_port() + 5  # avoid conflict with primary port
     srv.mcp_servers_json = mcp_json
     for k, v in kwargs.items():
         setattr(srv, k, v)
@@ -183,7 +183,7 @@ def test_mcp_tools_not_listed_when_not_configured():
     server = ServerPreset.router()
     server.server_tools = "all"
     server.no_ui = True
-    server.server_port = 8085
+    server.server_port = server_base_port() + 5
     server.start()
 
     try:
@@ -250,7 +250,7 @@ def test_mcp_tools_via_json_config_file():
         server = ServerPreset.router()
         server.server_tools = "all"
         server.no_ui = True
-        server.server_port = 8085
+        server.server_port = server_base_port() + 5
         server.mcp_servers_config = config_path
         server.start()
 
@@ -468,7 +468,7 @@ def test_mcp_config_file_errors():
     server = ServerPreset.router()
     server.server_tools = "all"
     server.no_ui = True
-    server.server_port = 8085
+    server.server_port = server_base_port() + 5
     server.mcp_servers_json = "not valid json"
     try:
         server.start()
@@ -480,7 +480,7 @@ def test_mcp_config_file_errors():
     server = ServerPreset.router()
     server.server_tools = "all"
     server.no_ui = True
-    server.server_port = 8085
+    server.server_port = server_base_port() + 5
     server.mcp_servers_config = "/nonexistent/path.json"
     try:
         server.start()

tools/server/tests/unit/test_tool_call.py

@@ -21,7 +21,7 @@ def create_server():
     global server
     server = ServerPreset.tinyllama2()
     server.model_alias = "tinyllama-2-tool-call"
-    server.server_port = 8081
+    server.server_port = server_base_port() + 1
     server.n_slots = 1
     server.n_ctx = 8192
     server.n_batch = 2048

tools/server/tests/utils.py

@@ -294,6 +294,7 @@ def start(self, timeout_seconds: int = DEFAULT_HTTP_TIMEOUT) -> None:
             server_args.append("--backend_sampling")
         if self.gcp_compat:
             env["AIP_MODE"] = "PREDICTION"
+            env["AIP_HTTP_PORT"] = str(self.server_port)
 
         args = [str(arg) for arg in [server_path, *server_args]]
         print(f"tests: starting server with: {' '.join(args)}")
@@ -518,6 +519,11 @@ def make_any_request(
 server_instances: Set[ServerProcess] = set()
 
 
+def server_base_port() -> int:
+    """Return the base port for this worker. Offset secondary servers from this value."""
+    return int(os.environ.get("BASE_PORT", "8080"))
+
+
 class ServerPreset:
     @staticmethod
     def load_all() -> None: