aipmodel 0.2.75__tar.gz → 0.2.77__tar.gz

{aipmodel-0.2.75 → aipmodel-0.2.77}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aipmodel
-Version: 0.2.75
+Version: 0.2.77
 Summary: SDK for model registration, versioning, and storage
 Home-page: https://github.com/AIP-MLOPS/model-registry
 Author: AIP MLOPS Team

{aipmodel-0.2.75 → aipmodel-0.2.77}/aipmodel/CephS3Manager.py

@@ -37,6 +37,11 @@ _TRANSIENT_TRANSFER_ERRORS = tuple(
         "EndpointConnectionError",
         "ReadTimeoutError",
         "IncompleteReadError",
+        # boto3 wraps a mid-stream connection break (urllib3 ProtocolError /
+        # http IncompleteRead) into ResponseStreamingError before it reaches the
+        # caller, so the underlying transient types below never match on their
+        # own for a download that dies mid-body. Retry on the wrapper too.
+        "ResponseStreamingError",
     )
     if hasattr(_botocore_exceptions, _name)
 ) + (
@@ -628,8 +633,16 @@ class CephS3Manager:
             except _TRANSIENT_TRANSFER_ERRORS as e:
                 attempt += 1
                 if attempt > self.max_transfer_retries:
-                    if self.verbose:
-                        print(f"[SDK_FAIL] {label}: transient error persisted after {self.max_transfer_retries} retries: {e}")
+                    # Always log this (not gated by verbose): a transient error that
+                    # survives every retry is a NETWORK problem, and the user must be
+                    # able to tell it apart from a Model Registry fault in the task log.
+                    print(
+                        f"[SDK_NETWORK_ERROR] {label}: NETWORK/connection failure — this is "
+                        f"NOT a Model Registry problem. The connection to the storage endpoint "
+                        f"kept dropping mid-transfer and did not recover after "
+                        f"{self.max_transfer_retries} automatic retries. Check network "
+                        f"stability and retry. Underlying: {type(e).__name__}: {e}"
+                    )
                     raise
                 delay = min(self.transfer_retry_backoff * (2 ** (attempt - 1)), self.transfer_retry_max_delay)
                 if self.verbose:

{aipmodel-0.2.75 → aipmodel-0.2.77}/aipmodel/__init__.py

@@ -11,5 +11,5 @@ if _os.environ.get("AIPMODEL_UPDATE_CHECK", "").lower() in ("1", "true", "yes"):
 
 __all__ = ["MLOpsManager", "CephS3Manager"]
 
-__version__ = "0.2.75"
+__version__ = "0.2.77"
 __description__ = "SDK for model registration, versioning, and storage"

{aipmodel-0.2.75 → aipmodel-0.2.77}/aipmodel/model_registry.py

@@ -20,6 +20,33 @@ from .CephS3Manager import CephS3Manager
 from .acl_manager import AclManager
 from .exceptions import ModelRegistryError, ModelNotFoundError, MetadataError, StorageError, IntegrityDisabledError
 
+
+def _is_transient_network_error(exc):
+    """
+    True when a failure was caused by a transient NETWORK/connection problem
+    (dropped connection, mid-stream read break, timeout) rather than a Model
+    Registry / storage-content fault. Walks the exception chain and also matches
+    the wrapped message, because transfer errors bubble up re-wrapped as
+    ValueError ("S3 Transfer failed: ... Connection broken: IncompleteRead ...").
+    Used only to add a clear, human-readable "this is the network, not the
+    registry" note to the task log — it never changes control flow.
+    """
+    _NET_MARKERS = (
+        "connection broken", "incompleteread", "responsestreaming",
+        "connection reset", "connection aborted", "connection refused",
+        "connection was closed", "timed out", "read timeout", "broken pipe",
+        "network is unreachable", "eof occurred", "endpointconnection",
+        "connectionerror", "name or service not known", "temporary failure in name resolution",
+    )
+    seen = set()
+    e = exc
+    while e is not None and id(e) not in seen:
+        seen.add(id(e))
+        if any(m in str(e).lower() for m in _NET_MARKERS):
+            return True
+        e = getattr(e, "__cause__", None) or getattr(e, "__context__", None)
+    return False
+
 load_dotenv()
 
 logger = logging.getLogger(__name__)
@@ -1102,6 +1129,17 @@ class MLOpsManager:
             return model_id
 
         except (Exception, KeyboardInterrupt) as e:
+            if _is_transient_network_error(e):
+                print("[SDK_NETWORK_ERROR] ============================================================")
+                print("[SDK_NETWORK_ERROR] Model add FAILED because of a NETWORK / connection problem")
+                print("[SDK_NETWORK_ERROR] while transferring the model files.")
+                print("[SDK_NETWORK_ERROR] This is NOT a Model Registry problem — the registry and the")
+                print("[SDK_NETWORK_ERROR] storage are healthy. The connection to the storage/source")
+                print("[SDK_NETWORK_ERROR] endpoint dropped and did not recover after the automatic")
+                print("[SDK_NETWORK_ERROR] retries. Nothing was left half-written (rolled back).")
+                print("[SDK_NETWORK_ERROR] Action: check network stability to the S3 source / Ceph,")
+                print("[SDK_NETWORK_ERROR] then retry the add.")
+                print("[SDK_NETWORK_ERROR] ============================================================")
             print(f"[SDK_ERROR] Add model failed: {e}")
             if "model_id" in locals():
                 try:

{aipmodel-0.2.75 → aipmodel-0.2.77}/aipmodel.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aipmodel
-Version: 0.2.75
+Version: 0.2.77
 Summary: SDK for model registration, versioning, and storage
 Home-page: https://github.com/AIP-MLOPS/model-registry
 Author: AIP MLOPS Team

{aipmodel-0.2.75 → aipmodel-0.2.77}/setup.py

@@ -31,7 +31,7 @@ version, description = read_meta()
 
 setup(
     name="aipmodel",
-    version="0.2.75",
+    version="0.2.77",
     description=description,
     author="AIP MLOPS Team",
     author_email="mohmmadweb@gmail.com",