Fix: Timeout HOT/OHSOME requests below gunicorn timeout

main/config.py

@@ -30,6 +30,11 @@ class Config:
     COMMUNITY_DASHBOARD_DOMAIN = typing.cast("URLParseResult", settings.COMMUNITY_DASHBOARD_DOMAIN)
     WEBSITE_DOMAIN = typing.cast("URLParseResult", settings.WEBSITE_DOMAIN)
 
+    # NOTE: Gunicorn worker timeout (misc/prod/run_web.sh) is 40s, keep external API
+    # timeouts a few seconds below that so requests fail with a clear error instead
+    # of the gunicorn worker being killed mid-request.
+    DEFAULT_EXTERNAL_API_TIMEOUT = 38
+
     # NOTE: We get AOI for validate from HOT tasking manager
     HOT_TASKING_MANAGER_PROJECT_API_LINK = "https://tasking-manager-production-api.hotosm.org/api/v2/"
 

project_types/validate/api_calls.py

@@ -17,6 +17,10 @@ class ValidateApiCallError(Exception):
     pass
 
 
+class ValidateApiCallTimeoutError(ValidateApiCallError):
+    pass
+
+
 def remove_troublesome_chars(string: str | None):
     """Remove chars that cause trouble when pushed into postgres."""
     if type(string) is not str:
@@ -176,16 +180,22 @@ def remove_noise_and_add_user_info(json: dict[str, Any]) -> dict[str, Any]:
     return json
 
 
-# fixme(frozenhelium): merge this function with `ohsome` and also add appropriate messages to raised exceptions
-def get_object_count_from_ohsome(area: str, ohsome_filter: PydanticLongText) -> int | None:
+def get_object_count_from_ohsome(
+    area: str,
+    ohsome_filter: PydanticLongText,
+    timeout: int = Config.DEFAULT_EXTERNAL_API_TIMEOUT,
+) -> int | None:
     url = Config.OHSOME_API_LINK + "elements/count"
     data = {"bpolys": area, "filter": ohsome_filter}
 
     logger.info("Target: %s", url)
     logger.info("Filter: %s", ohsome_filter)
 
-    # fixme(frozenhelium): use httpx for proper timeout
-    response = requests.post(url, data=data, timeout=100)
+    try:
+        response = requests.post(url, data=data, timeout=timeout)
+    except requests.exceptions.Timeout as e:
+        logger.warning("ohsome element count request timed out")
+        raise ValidateApiCallTimeoutError("OHSOME request timed out. Please try again.") from e
     if response.status_code != 200:
         logger.warning(
             "ohsome element count request failed: check for errors in filter or geometries",
@@ -210,16 +220,24 @@ def get_object_count_from_ohsome(area: str, ohsome_filter: PydanticLongText) ->
     return int(value)
 
 
-def ohsome(request: dict[str, Any], area: str, properties: str | None = None) -> dict[str, Any]:
+def ohsome(
+    request: dict[str, Any],
+    area: str,
+    properties: str | None = None,
+    timeout: int = Config.DEFAULT_EXTERNAL_API_TIMEOUT,
+) -> dict[str, Any]:
     """Request data from Ohsome API."""
     url = Config.OHSOME_API_LINK + request["endpoint"]
     data = {"bpolys": area, "filter": request["filter"]}
     if properties:
         data["properties"] = properties
     logger.info("Target: %s", url)
     logger.info("Filter: %s", request["filter"])
-    # FIXME(tnagorra): Need to check what the timeout should be
-    response = requests.post(url, data=data, timeout=100)
+    try:
+        response = requests.post(url, data=data, timeout=timeout)
+    except requests.exceptions.Timeout as e:
+        logger.warning("ohsome request timed out")
+        raise ValidateApiCallTimeoutError("OHSOME request timed out. Please try again.") from e
     if response.status_code != 200:
         logger.warning(
             "ohsome request failed: check for errors in filter or geometries",

project_types/validate/project.py

@@ -28,7 +28,12 @@
 from main.config import Config
 from project_types.base import project as base_project
 from project_types.tile_map_service.base.project import create_json_dump
-from project_types.validate.api_calls import ValidateApiCallError, get_object_count_from_ohsome, ohsome
+from project_types.validate.api_calls import (
+    ValidateApiCallError,
+    ValidateApiCallTimeoutError,
+    get_object_count_from_ohsome,
+    ohsome,
+)
 from utils import fields as custom_fields
 from utils.asset_types.models import AoiGeometryAssetProperty
 from utils.common import Grouping, clean_up_none_keys, to_groups
@@ -157,9 +162,12 @@ def _validate_object_count(count: int | None) -> int:
     def test_geojson_url_project(url: str):
         logger.info("Fetching object geojson from %s", url)
 
-        # FIXME(frozenhelium): use predefined timeout duration
-        # FIXME(tnagorra): handle timeout error
-        response = requests.get(url, timeout=500)
+        try:
+            response = requests.get(url, timeout=Config.DEFAULT_EXTERNAL_API_TIMEOUT)
+        except requests.exceptions.Timeout as e:
+            raise base_project.ValidationException(
+                f"Failed to fetch object geojson from {url}: request timed out",
+            ) from e
         if response.status_code != 200:
             raise base_project.ValidationException(
                 f"Failed to fetch object geojson from {url}",
@@ -207,6 +215,8 @@ def test_ohsome_objects_from_aoi_asset(
                 feature_collection.model_dump_json(),
                 ohsome_filter,
             )
+        except ValidateApiCallTimeoutError as e:
+            raise base_project.ValidationException(str(e)) from e
         except Exception as e:
             raise base_project.ValidationException("Failed to get object_count from ohsome") from e
 
@@ -223,7 +233,12 @@ def test_tasking_manager_project(
         hot_tm_url = f"{Config.HOT_TASKING_MANAGER_PROJECT_API_LINK}projects/{hot_tm_id}/queries/aoi/?as_file=false"
         logger.info("Fetching AOI geojson on HOT from %s", hot_tm_url)
 
-        aoi_result = requests.get(hot_tm_url, timeout=500)
+        try:
+            aoi_result = requests.get(hot_tm_url, timeout=Config.DEFAULT_EXTERNAL_API_TIMEOUT)
+        except requests.exceptions.Timeout as e:
+            raise base_project.ValidationException(
+                f"Failed to fetch AOI GeoJSON from HOT Tasking Manager for tm_id {hot_tm_id}: request timed out",
+            ) from e
         if aoi_result.status_code != 200:
             raise base_project.ValidationException(
                 f"Failed to fetch AOI GeoJSON from HOT Tasking Manager for tm_id {hot_tm_id}",
@@ -269,6 +284,8 @@ def test_tasking_manager_project(
                 feature_collection.model_dump_json(),
                 ohsome_filter,
             )
+        except ValidateApiCallTimeoutError as e:
+            raise base_project.ValidationException(str(e)) from e
         except Exception as e:
             raise base_project.ValidationException("Failed to get object_count from ohsome") from e
 
@@ -311,9 +328,13 @@ def _get_object_geometry_from_ohsome(self, geojson: dict[typing.Any, typing.Any]
                 feature_collection.model_dump_json(),
                 properties="tags, metadata",
             )
+        except ValidateApiCallTimeoutError as e:
+            raise base_project.ValidationException(str(e)) from e
         except ValidateApiCallError as e:
             # NOTE: Handles calls from OHSOME, OSMCHA and OSM
-            raise base_project.ValidationException("Failed to fetch data from OHSOME/OSMCHA/OSM") from e
+            raise base_project.ValidationException(
+                "Failed to fetch data from OHSOME/OSMCHA/OSM",
+            ) from e
         except requests.JSONDecodeError as e:
             # NOTE: Handles calls from OHSOME and OSMCHA
             # OSM responds in XML format