integration-tests: Improves fetching sample data

by also retrying corrupted archives, being more specific regarding
retryable exceptions and setting a maximum of retries.

integration-tests/fetch-corpora.py

@@ -25,12 +25,19 @@
 import io
 import os
 import tarfile
+import zlib
 from pathlib import Path
 from tempfile import TemporaryFile
 from typing import Final, NamedTuple
 
 from httpx import AsyncClient, HTTPError
-from tenacity import retry, wait_random_exponential
+from tenacity import (
+    retry,
+    retry_if_exception_type,
+    stop_after_attempt,
+    wait_incrementing,
+    wait_random_exponential,
+)
 
 import delb
 
@@ -245,18 +252,17 @@ class Archive(NamedTuple):
 http_client: Final = AsyncClient()
 
 
-@retry(wait=wait_random_exponential(multiplier=1, max=120))
-async def fetch_resource(url: str, destination: io.BufferedWriter) -> bool:
+@retry(
+    retry=retry_if_exception_type(HTTPError),
+    wait=wait_random_exponential(multiplier=1, max=120),
+    stop=stop_after_attempt(12),
+    reraise=True,
+)
+async def fetch_resource(url: str, destination: io.BufferedWriter):
     async with http_client.stream("GET", url, follow_redirects=True) as response:
-        try:
-            async for chunk in response.aiter_bytes():
-                destination.write(chunk)
-        except HTTPError as e:
-            print(f"Failed to fetch {url}: {e}")
-            return False
-        else:
-            print(f"Downloaded {url} to {destination.name}")
-            return True
+        async for chunk in response.aiter_bytes():
+            destination.write(chunk)
+        print(f"Downloaded {url} to {destination.name}")
 
 
 def make_archive_filter(archive_description: Archive) -> callable:
@@ -286,6 +292,12 @@ def _filter(member: tarfile.TarInfo, path: str) -> tarfile.TarInfo | None:
 # there's no business like business
 
 
+@retry(
+    retry=retry_if_exception_type((EOFError, zlib.error)),
+    wait=wait_incrementing(start=1, increment=0.5, max=5),
+    stop=stop_after_attempt(12),
+    reraise=True,
+)
 async def fetch_archive(archive_description: Archive):
     target_folder = CORPORA_PATH / archive_description.target_directory
     if SKIP_EXISTING and target_folder.exists():