Better catching of httpx errors and retrying them

allenai · Nov 21, 2024 · f8dcdf6 · f8dcdf6
1 parent d6a0013
commit f8dcdf6
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/pdelfin/beakerpipeline.py b/pdelfin/beakerpipeline.py
@@ -184,7 +184,7 @@ async def process_page(args, session: httpx.AsyncClient, worker_id: int, pdf_s3_
                 output_tokens=base_response_data["usage"].get("completion_tokens", 0),
                 is_fallback=False,
             )
-        except (httpx.TimeoutException, httpx.ConnectError, asyncio.TimeoutError) as e:
+        except (httpx.TransportError, asyncio.TimeoutError) as e:
             logger.warning(f"Client error on attempt {attempt} for {pdf_s3_path}-{page_num}: {e}")
 
             # Now we want to do exponential backoff, and not count this as an actual page retry

diff --git a/pdelfin/version.py b/pdelfin/version.py
@@ -2,7 +2,7 @@
 _MINOR = "1"
 # On main and in a nightly release the patch should be one ahead of the last
 # released build.
-_PATCH = "41"
+_PATCH = "42"
 # This is mainly for nightly builds which have the suffix ".dev$DATE". See
 # https://semver.org/#is-v123-a-semantic-version for the semantics.
 _SUFFIX = ""