fix(cli, studio): don't retry on non request/http errors

src/datachain/remote/studio.py

@@ -14,6 +14,7 @@
 from urllib.parse import urlparse, urlunparse
 
 import websockets
+from requests.exceptions import HTTPError, Timeout
 
 from datachain.config import Config
 from datachain.error import DataChainError
@@ -104,8 +105,8 @@ def _get_team(self) -> str:
             raise DataChainError(
                 "Studio team is not set. "
                 "Use `datachain auth team <team_name>` "
-                "or environment variable `DVC_STUDIO_TEAM` to set it."
-                "You can also set it in the config file as team under studio."
+                "or environment variable `DVC_STUDIO_TEAM` to set it. "
+                "You can also set `studio.team` in the config file."
             )
 
         return team
@@ -158,15 +159,14 @@ def _send_request_msgpack(
             message = content.get("message", "")
         return Response(response_data, ok, message)
 
-    @retry_with_backoff(retries=5)
+    @retry_with_backoff(retries=3, errors=(HTTPError, Timeout))
     def _send_request(
         self, route: str, data: dict[str, Any], method: Optional[str] = "POST"
     ) -> Response[Any]:
         """
         Function that communicate Studio API.
         It will raise an exception, and try to retry, if 5xx status code is
-        returned, or if ConnectionError or Timeout exceptions are thrown from
-        requests lib
+        returned, or if Timeout exceptions is thrown from the requests lib
         """
         import requests
 
@@ -188,7 +188,7 @@ def _send_request(
         )
         try:
             response.raise_for_status()
-        except requests.exceptions.HTTPError:
+        except HTTPError:
             if _is_server_error(response.status_code):
                 # going to retry
                 raise

src/datachain/utils.py

@@ -1,6 +1,7 @@
 import glob
 import io
 import json
+import logging
 import os
 import os.path as osp
 import random
@@ -25,6 +26,9 @@
     import pandas as pd
     from typing_extensions import Self
 
+
+logger = logging.getLogger("datachain")
+
 NUL = b"\0"
 TIME_ZERO = datetime.fromtimestamp(0, tz=timezone.utc)
 
@@ -271,19 +275,25 @@ def flatten(items):
             yield item
 
 
-def retry_with_backoff(retries=5, backoff_sec=1):
+def retry_with_backoff(retries=5, backoff_sec=1, errors=(Exception,)):
     def retry(f):
         def wrapper(*args, **kwargs):
             num_tried = 0
             while True:
                 try:
                     return f(*args, **kwargs)
-                except Exception:
+                except errors:
                     if num_tried == retries:
                         raise
                     sleep = (
                         backoff_sec * 2** num_tried + random.uniform(0, 1)  # noqa: S311
                     )
+                    logger.exception(
+                        "Error in %s, retrying in %ds, attempt %d",
+                        f.__name__,
+                        sleep,
+                        num_tried,
+                    )
                     time.sleep(sleep)
                     num_tried += 1