[PATCH] fetch2/gcp: treat GatewayTimeout as fetch failure

[PATCH] fetch2/gcp: treat GatewayTimeout as fetch failure

[ecordonnier]

From: Sebastian Muxel <smuxel@snap.com>

blob.exists() and blob.download_to_filename() can raise
google.api_core.exceptions.GatewayTimeout after the GCS client's own
retries. Uncaught, that escapes as a hard error instead of a normal
fetch/checkstatus failure (and blocks mirror fallback on download).

Catch GatewayTimeout in checkstatus() and download(), log a warning so
the timeout is visible, and raise FetchError.

AI-Generated: Cursor with Grok 4.5
Signed-off-by: Sebastian Muxel <smuxel@snap.com>
Signed-off-by: Etienne Cordonnier <ecordonnier@snap.com>
---
 lib/bb/fetch2/gcp.py | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)

diff --git a/lib/bb/fetch2/gcp.py b/lib/bb/fetch2/gcp.py
index 86546d40b..462b8e97a 100644
--- a/lib/bb/fetch2/gcp.py
+++ b/lib/bb/fetch2/gcp.py
@@ -57,7 +57,7 @@ class GCP(FetchMethod):
         Fetch urls using the GCP API.
         Assumes localpath was called first.
         """
-        from google.api_core.exceptions import NotFound
+        from google.api_core.exceptions import GatewayTimeout, NotFound
         logger.debug2(f"Trying to download gs://{ud.host}{ud.path} to {ud.localpath}")
         if self.gcp_client is None:
             self.get_gcp_client()
@@ -71,6 +71,13 @@ class GCP(FetchMethod):
             blob.download_to_filename(ud.localpath)
         except NotFound:
             raise FetchError("The GCP API threw a NotFound exception")
+        except GatewayTimeout as e:
+            # The GCS client already retries GatewayTimeout internally.
+            # Raise FetchError so mirror fallback can proceed.
+            logger.warning(
+                f"GCP API GatewayTimeout while downloading gs://{ud.host}{ud.path}: {e}"
+            )
+            raise FetchError(f"Transient GCP API GatewayTimeout for gs://{ud.host}{ud.path}")
 
         # Additional sanity checks copied from the wget class (although there
         # are no known issues which mean these are required, treat the GCP API
@@ -88,6 +95,8 @@ class GCP(FetchMethod):
         """
         Check the status of a URL.
         """
+        from google.api_core.exceptions import GatewayTimeout
+
         logger.debug2(f"Checking status of gs://{ud.host}{ud.path}")
         if self.gcp_client is None:
             self.get_gcp_client()
@@ -96,7 +105,18 @@ class GCP(FetchMethod):
 
         # Path sometimes has leading slash, so strip it
         path = ud.path.lstrip("/")
-        if self.gcp_client.bucket(ud.host).blob(path).exists() == False:
+        try:
+            exists = self.gcp_client.bucket(ud.host).blob(path).exists()
+        except GatewayTimeout as e:
+            # The GCS client already retries GatewayTimeout internally.
+            # Surface a normal checkstatus failure and warn so the timeout
+            # is visible to operators.
+            logger.warning(
+                f"GCP API GatewayTimeout while checking gs://{ud.host}{ud.path}; treating as unavailable: {e}"
+            )
+            raise FetchError(f"Transient GCP API GatewayTimeout for gs://{ud.host}{ud.path}")
+
+        if exists == False:
             raise FetchError(f"The GCP API reported that gs://{ud.host}{ud.path} does not exist")
         else:
             return True
-- 
2.43.0