Merge pull request #5890 from opsmill/pog-event-retry-IFC-1280

Retry event query to prefects events API on failures

Author: ogenstad

backend/infrahub/exceptions.py

@@ -49,6 +49,15 @@ def __init__(self, message: str) -> None:
         super().__init__(self.message)
 
 
+class ServiceUnavailableError(Error):
+    HTTP_CODE: int = 503
+    DESCRIPTION = "Service unavailable"
+
+    def __init__(self, message: str) -> None:
+        self.message = message
+        super().__init__(self.message)
+
+
 class LockError(Error):
     pass
 

backend/infrahub/task_manager/event.py

@@ -1,12 +1,14 @@
 from __future__ import annotations
 
+import asyncio
 from typing import TYPE_CHECKING, Any
 
 from prefect.client.orchestration import PrefectClient, get_client
 from prefect.events.schemas.events import Event as PrefectEventModel
 from pydantic import BaseModel, Field, TypeAdapter
 
 from infrahub.core.constants import GLOBAL_BRANCH_NAME
+from infrahub.exceptions import ServiceUnavailableError
 from infrahub.log import get_logger
 from infrahub.utils import get_nested_dict
 
@@ -216,8 +218,17 @@ async def query_events(
     ) -> PrefectEventResponse:
         body = {"limit": limit, "filter": filters.model_dump(mode="json", exclude_none=True), "offset": offset}
 
-        response = await client._client.post("/infrahub/events/filter", json=body)
-        response.raise_for_status()
+        # Retry due to https://github.com/PrefectHQ/prefect/issues/16299
+        for _ in range(1, 5):
+            response = await client._client.post("/infrahub/events/filter", json=body)
+            if response.status_code == 200:
+                break
+            await asyncio.sleep(0.1)
+
+        if response.status_code != 200:
+            raise ServiceUnavailableError(
+                message=f"Unable to query prefect due to invalid response from the server (status_code={response.status_code})"
+            )
         data: dict[str, Any] = response.json()
 
         return PrefectEventResponse(