Transient HubSpot failures are retried under the shared rate-limit policy 🟩

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

etl/hubspot/deal_contacts_client.py

@@ -6,6 +6,7 @@ from hubspot.crm.associations.v4 import AssociationSpec  # type: ignore[reportMi
 from hubspot.crm.contacts import SimplePublicObjectInputForCreate  # type: ignore[reportMissingTypeStubs]
 
 from etl.hubspot.errors import HubspotRequestError
+from etl.hubspot.retry import call_with_retry
 
 # HubSpot-defined deal -> contact association ("deal_to_contact" in v3 terms).
 DEAL_TO_CONTACT_ASSOCIATION_TYPE_ID = 3
@@ -49,9 +50,11 @@ class HubspotDealContactsClient:
 
     def create_contact(self, properties: Dict[str, str]) -> str:
         try:
-            contact = self._client.crm.contacts.basic_api.create(  # type: ignore[reportUnknownMemberType]
-                simple_public_object_input_for_create=SimplePublicObjectInputForCreate(
-                    properties=properties
+            contact = call_with_retry(
+                lambda: self._client.crm.contacts.basic_api.create(  # type: ignore[reportUnknownMemberType]
+                    simple_public_object_input_for_create=SimplePublicObjectInputForCreate(
+                        properties=properties
+                    )
                 )
             )
         except Exception as error:
@@ -61,17 +64,19 @@ class HubspotDealContactsClient:
 
     def associate_contact_to_deal(self, deal_id: str, contact_id: str) -> None:
         try:
-            self._client.crm.associations.v4.basic_api.create(  # type: ignore[reportUnknownMemberType]
-                object_type="deals",
-                object_id=deal_id,
-                to_object_type="contacts",
-                to_object_id=contact_id,
-                association_spec=[
-                    AssociationSpec(
-                        association_category="HUBSPOT_DEFINED",
-                        association_type_id=DEAL_TO_CONTACT_ASSOCIATION_TYPE_ID,
-                    )
-                ],
+            call_with_retry(
+                lambda: self._client.crm.associations.v4.basic_api.create(  # type: ignore[reportUnknownMemberType]
+                    object_type="deals",
+                    object_id=deal_id,
+                    to_object_type="contacts",
+                    to_object_id=contact_id,
+                    association_spec=[
+                        AssociationSpec(
+                            association_category="HUBSPOT_DEFINED",
+                            association_type_id=DEAL_TO_CONTACT_ASSOCIATION_TYPE_ID,
+                        )
+                    ],
+                )
             )
         except Exception as error:
             raise _scrubbed(error) from None

etl/hubspot/hubspotClient.py

@@ -1,7 +1,5 @@
 import os
-import time
 from enum import Enum
-from http import HTTPStatus
 from typing import Optional, cast, Callable, Any
 
 from hubspot.client import Client  # type: ignore[reportMissingTypeStubs]
@@ -32,6 +30,7 @@ from hubspot.crm.associations.v4.models import (  # type: ignore[reportMissingTy
 from backend.app.config import get_settings
 from etl.hubspot.company_data import CompanyData
 from etl.hubspot.project_data import ProjectData
+from etl.hubspot.retry import call_with_retry
 from utils.logger import setup_logger
 
 import mimetypes
@@ -87,40 +86,8 @@ class HubspotClient:
         # self.client
 
     def _call_with_retry(self, fn: Callable[[], Any], max_retries: int = 2) -> Any:
-        """
-        Call fn(), retrying up to max_retries times on 429 rate-limit errors
-        or transient 5xx server errors.
-        Waits the minimal amount: the remaining interval window reported by HubSpot headers.
-        Falls back to the full interval (10s) if headers are absent.
-
-        Note: each HubSpot sub-module (deals, companies, etc.) ships its own ApiException
-        class with no shared base beyond Exception, so we detect retryable statuses via duck-typing.
-        """
-        retryable_statuses = {
-            HTTPStatus.TOO_MANY_REQUESTS,
-            HTTPStatus.INTERNAL_SERVER_ERROR,
-            HTTPStatus.BAD_GATEWAY,
-            HTTPStatus.SERVICE_UNAVAILABLE,
-            HTTPStatus.GATEWAY_TIMEOUT,
-        }
-        for attempt in range(max_retries + 1):
-            try:
-                return fn()
-            except Exception as e:
-                status = getattr(e, "status", None)
-                if status not in retryable_statuses or attempt == max_retries:
-                    raise
-                headers = getattr(e, "headers", None) or {}
-                interval_ms = int(
-                    headers.get("x-hubspot-ratelimit-interval-milliseconds", 10000)
-                )
-                wait_s = interval_ms / 1000.0
-                self.logger.warning(
-                    f"HubSpot {status} (attempt {attempt + 1}/{max_retries}), "
-                    f"waiting {wait_s:.1f}s before retry."
-                )
-                time.sleep(wait_s)
-        raise RuntimeError("Unreachable")  # pragma: no cover
+        """Delegates to the shared etl.hubspot.retry policy."""
+        return call_with_retry(fn, max_retries=max_retries, logger=self.logger)
 
     def get_deal_ids_from_company(self, company_id: str) -> list[str]:
         associations_api: AssociationsBasicApi = (  # type: ignore[reportUnknownMemberType]

etl/hubspot/retry.py

@@ -0,0 +1,51 @@
+import time
+from http import HTTPStatus
+from logging import Logger
+from typing import Callable, Optional, TypeVar
+
+from utils.logger import setup_logger
+
+T = TypeVar("T")
+
+RETRYABLE_STATUSES = {
+    HTTPStatus.TOO_MANY_REQUESTS,
+    HTTPStatus.INTERNAL_SERVER_ERROR,
+    HTTPStatus.BAD_GATEWAY,
+    HTTPStatus.SERVICE_UNAVAILABLE,
+    HTTPStatus.GATEWAY_TIMEOUT,
+}
+
+
+def call_with_retry(
+    fn: Callable[[], T],
+    max_retries: int = 2,
+    logger: Optional[Logger] = None,
+) -> T:
+    """
+    Call fn(), retrying up to max_retries times on 429 rate-limit errors
+    or transient 5xx server errors.
+    Waits the minimal amount: the remaining interval window reported by HubSpot headers.
+    Falls back to the full interval (10s) if headers are absent.
+
+    Note: each HubSpot sub-module (deals, companies, etc.) ships its own ApiException
+    class with no shared base beyond Exception, so we detect retryable statuses via duck-typing.
+    """
+    logger = logger or setup_logger()
+    for attempt in range(max_retries + 1):
+        try:
+            return fn()
+        except Exception as e:
+            status = getattr(e, "status", None)
+            if status not in RETRYABLE_STATUSES or attempt == max_retries:
+                raise
+            headers = getattr(e, "headers", None) or {}
+            interval_ms = int(
+                headers.get("x-hubspot-ratelimit-interval-milliseconds", 10000)
+            )
+            wait_s = interval_ms / 1000.0
+            logger.warning(
+                f"HubSpot {status} (attempt {attempt + 1}/{max_retries}), "
+                f"waiting {wait_s:.1f}s before retry."
+            )
+            time.sleep(wait_s)
+    raise RuntimeError("Unreachable")  # pragma: no cover