fix: add HTTP request timeouts to prevent hung threads

No timeout was set on any session.get/post/delete call, causing
threads to block indefinitely on slow or stalled connections.
Add connect_timeout (9.05s) and read_timeout (300s) to config
and wire them into _send_request. Also add Timeout to the
retryable exception set.

Signed-off-by: suhr25 <suhridmarwah07@gmail.com>

Author: suhr25

openml/_api_calls.py

@@ -370,16 +370,19 @@ def _send_request(  # noqa: C901, PLR0912
     # Error to raise in case of retrying too often. Will be set to the last observed exception.
     retry_raise_e: Exception | None = None
 
+    timeout = (config.connect_timeout, config.read_timeout)
     with requests.Session() as session:
         # Start at one to have a non-zero multiplier for the sleep
         for retry_counter in range(1, n_retries + 1):
             try:
                 if request_method == "get":
-                    response = session.get(url, params=data, headers=_HEADERS)
+                    response = session.get(url, params=data, headers=_HEADERS, timeout=timeout)
                 elif request_method == "delete":
-                    response = session.delete(url, params=data, headers=_HEADERS)
+                    response = session.delete(url, params=data, headers=_HEADERS, timeout=timeout)
                 elif request_method == "post":
-                    response = session.post(url, data=data, files=files, headers=_HEADERS)
+                    response = session.post(
+                        url, data=data, files=files, headers=_HEADERS, timeout=timeout
+                    )
                 else:
                     raise NotImplementedError()
 
@@ -424,6 +427,7 @@ def _send_request(  # noqa: C901, PLR0912
                     ) from e
                 retry_raise_e = e
             except (
+                requests.exceptions.Timeout,
                 requests.exceptions.ChunkedEncodingError,
                 requests.exceptions.ConnectionError,
                 requests.exceptions.SSLError,

openml/config.py

@@ -37,6 +37,8 @@ class _Config(TypedDict):
     retry_policy: Literal["human", "robot"]
     connection_n_retries: int
     show_progress: bool
+    connect_timeout: float
+    read_timeout: float
 
 
 def _create_log_handlers(create_file_handler: bool = True) -> None:  # noqa: FBT002
@@ -157,6 +159,8 @@ def _resolve_default_cache_dir() -> Path:
     "retry_policy": "human",
     "connection_n_retries": 5,
     "show_progress": False,
+    "connect_timeout": 9.05,
+    "read_timeout": 300.0,
 }
 
 # Default values are actually added here in the _setup() function which is
@@ -186,6 +190,8 @@ def get_server_base_url() -> str:
 
 retry_policy: Literal["human", "robot"] = _defaults["retry_policy"]
 connection_n_retries: int = _defaults["connection_n_retries"]
+connect_timeout: float = _defaults["connect_timeout"]
+read_timeout: float = _defaults["read_timeout"]
 
 
 def set_retry_policy(value: Literal["human", "robot"], n_retries: int | None = None) -> None:
@@ -343,6 +349,8 @@ def _setup(config: _Config | None = None) -> None:
     global _root_cache_directory  # noqa: PLW0603
     global avoid_duplicate_runs  # noqa: PLW0603
     global show_progress  # noqa: PLW0603
+    global connect_timeout  # noqa: PLW0603
+    global read_timeout  # noqa: PLW0603
 
     config_file = determine_config_file_path()
     config_dir = config_file.parent
@@ -364,6 +372,8 @@ def _setup(config: _Config | None = None) -> None:
     apikey = config["apikey"]
     server = config["server"]
     show_progress = config["show_progress"]
+    connect_timeout = float(config.get("connect_timeout", _defaults["connect_timeout"]))  # type: ignore[union-attr]
+    read_timeout = float(config.get("read_timeout", _defaults["read_timeout"]))  # type: ignore[union-attr]
     n_retries = int(config["connection_n_retries"])
 
     set_retry_policy(config["retry_policy"], n_retries)
@@ -445,6 +455,8 @@ def get_config_as_dict() -> _Config:
         "connection_n_retries": connection_n_retries,
         "retry_policy": retry_policy,
         "show_progress": show_progress,
+        "connect_timeout": connect_timeout,
+        "read_timeout": read_timeout,
     }
 
 

tests/test_openml/test_config.py

@@ -84,8 +84,10 @@ def test_get_config_as_dict(self):
         _config["connection_n_retries"] = 20
         _config["retry_policy"] = "robot"
         _config["show_progress"] = False
+        _config["connect_timeout"] = openml.config.connect_timeout
+        _config["read_timeout"] = openml.config.read_timeout
         assert isinstance(config, dict)
-        assert len(config) == 7
+        assert len(config) == 9
         self.assertDictEqual(config, _config)
 
     def test_setup_with_config(self):
@@ -98,6 +100,8 @@ def test_setup_with_config(self):
         _config["retry_policy"] = "human"
         _config["connection_n_retries"] = 100
         _config["show_progress"] = False
+        _config["connect_timeout"] = 5.0
+        _config["read_timeout"] = 120.0
         orig_config = openml.config.get_config_as_dict()
         openml.config._setup(_config)
         updated_config = openml.config.get_config_as_dict()