mlrun 1.7.0rc6__py3-none-any.whl → 1.7.0rc8__py3-none-any.whl

mlrun/__main__.py

@@ -505,6 +505,8 @@ def build(
         if kfp:
             print("Runtime:")
             pprint(runtime)
+        # use kind = "job" by default if not specified
+        runtime.setdefault("kind", "job")
         func = new_function(runtime=runtime)
 
     elif func_url:

mlrun/common/constants.py

@@ -14,4 +14,10 @@
 #
 IMAGE_NAME_ENRICH_REGISTRY_PREFIX = "."  # prefix for image name to enrich with registry
 MLRUN_CREATED_LABEL = "mlrun-created"
+MLRUN_MODEL_CONF = "model-conf"
+MLRUN_SERVING_SPEC_MOUNT_PATH = f"/tmp/mlrun/{MLRUN_MODEL_CONF}"
+MLRUN_SERVING_SPEC_FILENAME = "serving_spec.json"
+MLRUN_SERVING_SPEC_PATH = (
+    f"{MLRUN_SERVING_SPEC_MOUNT_PATH}/{MLRUN_SERVING_SPEC_FILENAME}"
+)
 MYSQL_MEDIUMBLOB_SIZE_BYTES = 16 * 1024 * 1024

mlrun/common/schemas/__init__.py

@@ -21,6 +21,7 @@ from .api_gateway import (
     APIGatewayMetadata,
     APIGatewaysOutput,
     APIGatewaySpec,
+    APIGatewayState,
     APIGatewayStatus,
     APIGatewayUpstream,
 )
@@ -124,6 +125,7 @@ from .model_monitoring import (
     EventFieldType,
     EventKeyMetrics,
     Features,
+    FeatureSetFeatures,
     FeatureValues,
     GrafanaColumn,
     GrafanaDataPoint,
@@ -139,6 +141,7 @@ from .model_monitoring import (
     ModelMonitoringMode,
     ModelMonitoringStoreKinds,
     MonitoringFunctionNames,
+    PrometheusEndpoints,
     TimeSeriesTarget,
 )
 from .notification import (

mlrun/common/schemas/api_gateway.py

@@ -36,6 +36,13 @@ class APIGatewayAuthenticationMode(mlrun.common.types.StrEnum):
             )
 
 
+class APIGatewayState(mlrun.common.types.StrEnum):
+    none = ""
+    ready = "ready"
+    error = "error"
+    waiting_for_provisioning = "waitingForProvisioning"
+
+
 class _APIGatewayBaseModel(pydantic.BaseModel):
     class Config:
         extra = pydantic.Extra.allow
@@ -72,7 +79,7 @@ class APIGatewaySpec(_APIGatewayBaseModel):
 
 class APIGatewayStatus(_APIGatewayBaseModel):
     name: Optional[str]
-    state: Optional[str]
+    state: Optional[APIGatewayState]
 
 
 class APIGateway(_APIGatewayBaseModel):

mlrun/common/schemas/model_monitoring/__init__.py

@@ -22,6 +22,7 @@ from .constants import (
     EventFieldType,
     EventKeyMetrics,
     EventLiveStats,
+    FeatureSetFeatures,
     FileTargetKind,
     FunctionURI,
     ModelEndpointTarget,
@@ -29,9 +30,12 @@ from .constants import (
     ModelMonitoringStoreKinds,
     MonitoringFunctionNames,
     ProjectSecretKeys,
+    PrometheusEndpoints,
     PrometheusMetric,
+    SchedulingKeys,
     TimeSeriesTarget,
     VersionedModel,
+    WriterEvent,
 )
 from .grafana import (
     GrafanaColumn,

mlrun/common/schemas/model_monitoring/constants.py

@@ -21,6 +21,12 @@ import mlrun.common.helpers
 from mlrun.common.types import StrEnum
 
 
+class MonitoringStrEnum(StrEnum):
+    @classmethod
+    def list(cls):
+        return list(map(lambda c: c.value, cls))
+
+
 class EventFieldType:
     FUNCTION_URI = "function_uri"
     FUNCTION = "function"
@@ -77,6 +83,20 @@ class EventFieldType:
     SAMPLE_PARQUET_PATH = "sample_parquet_path"
 
 
+class FeatureSetFeatures(MonitoringStrEnum):
+    LATENCY = EventFieldType.LATENCY
+    ERROR_COUNT = EventFieldType.ERROR_COUNT
+    METRICS = EventFieldType.METRICS
+
+    @classmethod
+    def time_stamp(cls):
+        return EventFieldType.TIMESTAMP
+
+    @classmethod
+    def entity(cls):
+        return EventFieldType.ENDPOINT_ID
+
+
 class ApplicationEvent:
     APPLICATION_NAME = "application_name"
     CURRENT_STATS = "current_stats"
@@ -89,7 +109,7 @@ class ApplicationEvent:
     OUTPUT_STREAM_URI = "output_stream_uri"
 
 
-class WriterEvent(StrEnum):
+class WriterEvent(MonitoringStrEnum):
     APPLICATION_NAME = "application_name"
     ENDPOINT_ID = "endpoint_id"
     START_INFER_TIME = "start_infer_time"
@@ -101,10 +121,6 @@ class WriterEvent(StrEnum):
     RESULT_EXTRA_DATA = "result_extra_data"
     CURRENT_STATS = "current_stats"
 
-    @classmethod
-    def list(cls):
-        return list(map(lambda c: c.value, cls))
-
 
 class EventLiveStats:
     LATENCY_AVG_5M = "latency_avg_5m"
@@ -146,6 +162,9 @@ class ModelMonitoringStoreKinds:
 
 class SchedulingKeys:
     LAST_ANALYZED = "last_analyzed"
+    ENDPOINT_ID = "endpoint_id"
+    APPLICATION_NAME = "application_name"
+    UID = "uid"
 
 
 class FileTargetKind:
@@ -155,6 +174,8 @@ class FileTargetKind:
     PARQUET = "parquet"
     APPS_PARQUET = "apps_parquet"
     LOG_STREAM = "log_stream"
+    APP_RESULTS = "app_results"
+    MONITORING_SCHEDULES = "monitoring_schedules"
 
 
 class ModelMonitoringMode(str, Enum):
@@ -177,20 +198,16 @@ class PrometheusMetric:
     DRIFT_STATUS = "drift_status"
 
 
-class MonitoringFunctionNames:
-    WRITER = "model-monitoring-writer"
-    BATCH = "model-monitoring-batch"
-    APPLICATION_CONTROLLER = "model-monitoring-controller"
-    STREAM = "model-monitoring-stream"
+class PrometheusEndpoints(MonitoringStrEnum):
+    MODEL_MONITORING_METRICS = "/model-monitoring-metrics"
+    MONITORING_BATCH_METRICS = "/monitoring-batch-metrics"
+    MONITORING_DRIFT_STATUS = "/monitoring-drift-status"
+
 
-    @staticmethod
-    def all():
-        return [
-            MonitoringFunctionNames.WRITER,
-            MonitoringFunctionNames.STREAM,
-            MonitoringFunctionNames.BATCH,
-            MonitoringFunctionNames.APPLICATION_CONTROLLER,
-        ]
+class MonitoringFunctionNames(MonitoringStrEnum):
+    STREAM = "model-monitoring-stream"
+    APPLICATION_CONTROLLER = "model-monitoring-controller"
+    WRITER = "model-monitoring-writer"
 
 
 @dataclass

mlrun/common/schemas/project.py

@@ -87,6 +87,7 @@ class ProjectSpec(pydantic.BaseModel):
     custom_packagers: typing.Optional[list[tuple[str, bool]]] = None
     default_image: typing.Optional[str] = None
     build: typing.Optional[ImageBuilder] = None
+    default_function_node_selector: typing.Optional[dict] = {}
 
     class Config:
         extra = pydantic.Extra.allow

mlrun/common/types.py

@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-#
 
 import enum
 
@@ -23,3 +22,10 @@ class StrEnum(str, enum.Enum):
 
     def __repr__(self):
         return self.value
+
+
+# Partial backport from Python 3.11
+# https://docs.python.org/3/library/http.html#http.HTTPMethod
+class HTTPMethod(StrEnum):
+    GET = "GET"
+    POST = "POST"

mlrun/config.py

@@ -362,6 +362,8 @@ default_config = {
             # - mlrun.runtimes.nuclio.function.enrich_function_with_ingress
             "add_templated_ingress_host_mode": "never",
             "explicit_ack": "enabled",
+            # size of serving spec to move to config maps
+            "serving_spec_env_cutoff": 4096,
         },
         "logs": {
             "decode": {
@@ -479,6 +481,14 @@ default_config = {
             # if set to true, will log a warning for trying to use run db functionality while in nop db mode
             "verbose": True,
         },
+        "pagination": {
+            "default_page_size": 20,
+            "pagination_cache": {
+                "interval": 60,
+                "ttl": 3600,
+                "max_size": 10000,
+            },
+        },
     },
     "model_endpoint_monitoring": {
         "serving_stream_args": {"shard_count": 1, "retention_period_hours": 24},
@@ -498,10 +508,9 @@ default_config = {
         # when the user is working in CE environment and has not provided any stream path.
         "default_http_sink": "http://nuclio-{project}-model-monitoring-stream.{namespace}.svc.cluster.local:8080",
         "default_http_sink_app": "http://nuclio-{project}-{application_name}.{namespace}.svc.cluster.local:8080",
-        "batch_processing_function_branch": "master",
         "parquet_batching_max_events": 10_000,
         "parquet_batching_timeout_secs": timedelta(minutes=1).total_seconds(),
-        # See mlrun.model_monitoring.stores.ModelEndpointStoreType for available options
+        # See mlrun.model_monitoring.db.stores.ObjectStoreFactory for available options
         "store_type": "v3io-nosql",
         "endpoint_store_connection": "",
     },
@@ -542,6 +551,7 @@ default_config = {
             "nosql": "v3io:///projects/{project}/FeatureStore/{name}/{kind}",
             # "authority" is optional and generalizes [userinfo "@"] host [":" port]
             "redisnosql": "redis://{authority}/projects/{project}/FeatureStore/{name}/{kind}",
+            "dsnosql": "ds://{ds_profile_name}/projects/{project}/FeatureStore/{name}/{kind}",
         },
         "default_targets": "parquet,nosql",
         "default_job_image": "mlrun/mlrun",
@@ -616,8 +626,9 @@ default_config = {
     },
     "workflows": {
         "default_workflow_runner_name": "workflow-runner-{}",
-        # Default timeout seconds for retrieving workflow id after execution:
-        "timeouts": {"local": 120, "kfp": 30, "remote": 90},
+        # Default timeout seconds for retrieving workflow id after execution
+        # Remote workflow timeout is the maximum between remote and the inner engine timeout
+        "timeouts": {"local": 120, "kfp": 60, "remote": 60 * 5},
     },
     "log_collector": {
         "address": "localhost:8282",
@@ -1066,7 +1077,7 @@ class Config:
         target: str = "online",
         artifact_path: str = None,
         function_name: str = None,
-    ) -> str:
+    ) -> typing.Union[str, list[str]]:
         """Get the full path from the configuration based on the provided project and kind.
 
         :param project:         Project name.
@@ -1082,7 +1093,8 @@ class Config:
                                 relative artifact path will be taken from the global MLRun artifact path.
         :param function_name:    Application name, None for model_monitoring_stream.
 
-        :return:                Full configured path for the provided kind.
+        :return:                Full configured path for the provided kind. Can be either a single path
+                                or a list of paths in the case of the online model monitoring stream path.
         """
 
         if target != "offline":
@@ -1104,10 +1116,22 @@ class Config:
                     if function_name is None
                     else f"{kind}-{function_name.lower()}",
                 )
-            return mlrun.mlconf.model_endpoint_monitoring.store_prefixes.default.format(
-                project=project,
-                kind=kind,
-            )
+            elif kind == "stream":  # return list for mlrun<1.6.3 BC
+                return [
+                    mlrun.mlconf.model_endpoint_monitoring.store_prefixes.default.format(
+                        project=project,
+                        kind=kind,
+                    ),  # old stream uri (pipelines) for BC ML-6043
+                    mlrun.mlconf.model_endpoint_monitoring.store_prefixes.user_space.format(
+                        project=project,
+                        kind=kind,
+                    ),  # new stream uri (projects)
+                ]
+            else:
+                return mlrun.mlconf.model_endpoint_monitoring.store_prefixes.default.format(
+                    project=project,
+                    kind=kind,
+                )
 
         # Get the current offline path from the configuration
         file_path = mlrun.mlconf.model_endpoint_monitoring.offline_storage_path.format(

mlrun/data_types/data_types.py

@@ -41,6 +41,7 @@ class ValueType(str, Enum):
     BYTES = "bytes"
     STRING = "str"
     DATETIME = "datetime"
+    LIST = "List"
     BYTES_LIST = "List[bytes]"
     STRING_LIST = "List[string]"
     INT32_LIST = "List[int32]"
@@ -48,6 +49,7 @@ class ValueType(str, Enum):
     DOUBLE_LIST = "List[float]"
     FLOAT_LIST = "List[float32]"
     BOOL_LIST = "List[bool]"
+    Tuple = "Tuple"
 
 
 def pd_schema_to_value_type(value):
@@ -102,6 +104,8 @@ def python_type_to_value_type(value_type):
         "datetime64[ns]": ValueType.INT64,
         "datetime64[ns, tz]": ValueType.INT64,
         "category": ValueType.STRING,
+        "list": ValueType.LIST,
+        "tuple": ValueType.Tuple,
     }
 
     if type_name in type_map:

mlrun/datastore/alibaba_oss.py

@@ -0,0 +1,130 @@
+# Copyright 2023 Iguazio
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import time
+from datetime import datetime
+from pathlib import Path
+from urllib.parse import urlparse
+
+import oss2
+from fsspec.registry import get_filesystem_class
+
+import mlrun.errors
+
+from .base import DataStore, FileStats, makeDatastoreSchemaSanitizer
+
+
+class OSSStore(DataStore):
+    using_bucket = True
+
+    def __init__(self, parent, schema, name, endpoint="", secrets: dict = None):
+        super().__init__(parent, name, schema, endpoint, secrets)
+        # will be used in case user asks to assume a role and work through fsspec
+
+        access_key_id = self._get_secret_or_env("ALIBABA_ACCESS_KEY_ID")
+        secret_key = self._get_secret_or_env("ALIBABA_SECRET_ACCESS_KEY")
+        endpoint_url = self._get_secret_or_env("ALIBABA_ENDPOINT_URL")
+        if access_key_id and secret_key and endpoint_url:
+            self.auth = oss2.Auth(access_key_id, secret_key)
+            self.endpoint_url = endpoint_url
+        else:
+            raise mlrun.errors.MLRunInvalidArgumentError(
+                "missing ALIBABA_ACCESS_KEY_ID or ALIBABA_SECRET_ACCESS_KEY ALIBABA_ENDPOINT_URL in environment"
+            )
+
+    @property
+    def filesystem(self):
+        """return fsspec file system object, if supported"""
+        if self._filesystem:
+            return self._filesystem
+        try:
+            import ossfs  # noqa
+        except ImportError as exc:
+            raise ImportError("ALIBABA ossfs not installed") from exc
+        filesystem_class = get_filesystem_class(protocol=self.kind)
+        self._filesystem = makeDatastoreSchemaSanitizer(
+            filesystem_class,
+            using_bucket=self.using_bucket,
+            **self.get_storage_options(),
+        )
+        return self._filesystem
+
+    def get_storage_options(self):
+        res = dict(
+            endpoint=self._get_secret_or_env("ALIBABA_ENDPOINT_URL"),
+            key=self._get_secret_or_env("ALIBABA_ACCESS_KEY_ID"),
+            secret=self._get_secret_or_env("ALIBABA_SECRET_ACCESS_KEY"),
+        )
+        return self._sanitize_storage_options(res)
+
+    def get_bucket_and_key(self, key):
+        path = self._join(key)[1:]
+        return self.endpoint, path
+
+    def upload(self, key, src_path):
+        bucket, key = self.get_bucket_and_key(key)
+        oss = oss2.Bucket(self.auth, self.endpoint_url, bucket)
+        oss.put_object(key, open(src_path, "rb"))
+
+    def get(self, key, size=None, offset=0):
+        bucket, key = self.get_bucket_and_key(key)
+        oss = oss2.Bucket(self.auth, self.endpoint_url, bucket)
+        if size or offset:
+            return oss.get_object(key, byte_range=self.get_range(size, offset)).read()
+        return oss.get_object(key).read()
+
+    def put(self, key, data, append=False):
+        bucket, key = self.get_bucket_and_key(key)
+        oss = oss2.Bucket(self.auth, self.endpoint_url, bucket)
+        oss.put_object(key, data)
+
+    def stat(self, key):
+        bucket, key = self.get_bucket_and_key(key)
+        oss = oss2.Bucket(self.auth, self.endpoint_url, bucket)
+        obj = oss.get_object_meta(key)
+        size = obj.content_length
+        modified = datetime.fromtimestamp(obj.last_modified)
+        return FileStats(size, time.mktime(modified.timetuple()))
+
+    def listdir(self, key):
+        remote_path = self._convert_key_to_remote_path(key)
+        if self.filesystem.isfile(remote_path):
+            return key
+        remote_path = f"{remote_path}/**"
+        files = self.filesystem.glob(remote_path)
+        key_length = len(key)
+        files = [
+            f.split("/", 1)[1][key_length:] for f in files if len(f.split("/")) > 1
+        ]
+        return files
+
+    def delete(self, key):
+        bucket, key = self.get_bucket_and_key(key)
+        oss = oss2.Bucket(self.auth, self.endpoint_url, bucket)
+        oss.delete_object(key)
+
+    def _convert_key_to_remote_path(self, key):
+        key = key.strip("/")
+        schema = urlparse(key).scheme
+        #  if called without passing dataitem - like in fset.purge_targets,
+        #  key will include schema.
+        if not schema:
+            key = Path(self.endpoint, key).as_posix()
+        return key
+
+    @staticmethod
+    def get_range(size, offset):
+        if size:
+            return [offset, size]
+        return [offset, None]

mlrun/datastore/azure_blob.py

@@ -158,18 +158,17 @@ class AzureBlobStore(DataStore):
                     st[key] = parsed_value
 
         account_name = st.get("account_name")
-        if not account_name:
-            raise mlrun.errors.MLRunInvalidArgumentError(
-                "Property 'account_name' is absent both in storage settings and connection string"
-            )
         if primary_url:
             if primary_url.startswith("http://"):
                 primary_url = primary_url[len("http://") :]
             if primary_url.startswith("https://"):
                 primary_url = primary_url[len("https://") :]
             host = primary_url
-        else:
+        elif account_name:
             host = f"{account_name}.{service}.core.windows.net"
+        else:
+            return res
+
         if "account_key" in st:
             res[f"spark.hadoop.fs.azure.account.key.{host}"] = st["account_key"]
 

mlrun/datastore/base.py

@@ -27,6 +27,7 @@ import requests
 import urllib3
 from deprecated import deprecated
 
+import mlrun.config
 import mlrun.errors
 from mlrun.errors import err_to_str
 from mlrun.utils import StorePrefix, is_ipython, logger
@@ -34,10 +35,6 @@ from mlrun.utils import StorePrefix, is_ipython, logger
 from .store_resources import is_store_uri, parse_store_uri
 from .utils import filter_df_start_end_time, select_columns_from_df
 
-verify_ssl = False
-if not verify_ssl:
-    urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
-
 
 class FileStats:
     def __init__(self, size, modified, content_type=None):
@@ -633,17 +630,6 @@ def basic_auth_header(user, password):
     return {"Authorization": authstr}
 
 
-def http_get(url, headers=None, auth=None):
-    try:
-        response = requests.get(url, headers=headers, auth=auth, verify=verify_ssl)
-    except OSError as exc:
-        raise OSError(f"error: cannot connect to {url}: {err_to_str(exc)}")
-
-    mlrun.errors.raise_for_status(response)
-
-    return response.content
-
-
 class HttpStore(DataStore):
     def __init__(self, parent, schema, name, endpoint="", secrets: dict = None):
         super().__init__(parent, name, schema, endpoint, secrets)
@@ -671,7 +657,7 @@ class HttpStore(DataStore):
         raise ValueError("unimplemented")
 
     def get(self, key, size=None, offset=0):
-        data = http_get(self.url + self._join(key), self._headers, self.auth)
+        data = self._http_get(self.url + self._join(key), self._headers, self.auth)
         if offset:
             data = data[offset:]
         if size:
@@ -691,6 +677,26 @@ class HttpStore(DataStore):
                 f"schema as it is not secure and is not recommended."
             )
 
+    def _http_get(
+        self,
+        url,
+        headers=None,
+        auth=None,
+    ):
+        # import here to prevent import cycle
+        from mlrun.config import config as mlconf
+
+        verify_ssl = mlconf.httpdb.http.verify
+        try:
+            if not verify_ssl:
+                urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
+            response = requests.get(url, headers=headers, auth=auth, verify=verify_ssl)
+        except OSError as exc:
+            raise OSError(f"error: cannot connect to {url}: {err_to_str(exc)}")
+
+        mlrun.errors.raise_for_status(response)
+        return response.content
+
 
 # This wrapper class is designed to extract the 'ds' schema and profile name from URL-formatted paths.
 # Within fsspec, the AbstractFileSystem::_strip_protocol() internal method is used to handle complete URL paths.

mlrun/datastore/datastore.py

@@ -98,6 +98,10 @@ def schema_to_store(schema):
         from .hdfs import HdfsStore
 
         return HdfsStore
+    elif schema == "oss":
+        from .alibaba_oss import OSSStore
+
+        return OSSStore
     else:
         raise ValueError(f"unsupported store scheme ({schema})")
 

mlrun/datastore/datastore_profile.py

@@ -68,6 +68,9 @@ class TemporaryClientDatastoreProfiles(metaclass=mlrun.utils.singleton.Singleton
     def get(self, key):
         return self._data.get(key, None)
 
+    def remove(self, key):
+        self._data.pop(key, None)
+
 
 class DatastoreProfileBasic(DatastoreProfile):
     type: str = pydantic.Field("basic")
@@ -460,3 +463,7 @@ def register_temporary_client_datastore_profile(profile: DatastoreProfile):
     It's beneficial for testing purposes.
     """
     TemporaryClientDatastoreProfiles().add(profile)
+
+
+def remove_temporary_client_datastore_profile(profile_name: str):
+    TemporaryClientDatastoreProfiles().remove(profile_name)

mlrun/datastore/google_cloud_storage.py

@@ -132,7 +132,7 @@ class GoogleCloudStorageStore(DataStore):
         self.filesystem.rm(path=path, recursive=recursive, maxdepth=maxdepth)
 
     def get_spark_options(self):
-        res = None
+        res = {}
         st = self.get_storage_options()
         if "token" in st:
             res = {"spark.hadoop.google.cloud.auth.service.account.enable": "true"}

mlrun/datastore/sources.py

@@ -854,12 +854,11 @@ class StreamSource(OnlineSource):
         super().__init__(name, attributes=attrs, **kwargs)
 
     def add_nuclio_trigger(self, function):
-        store, path, url = mlrun.store_manager.get_or_create_store(self.path)
+        store, _, url = mlrun.store_manager.get_or_create_store(self.path)
         if store.kind != "v3io":
             raise mlrun.errors.MLRunInvalidArgumentError(
                 "Only profiles that reference the v3io datastore can be used with StreamSource"
             )
-        path = "v3io:/" + path
         storage_options = store.get_storage_options()
         access_key = storage_options.get("v3io_access_key")
         endpoint, stream_path = parse_path(url)
@@ -883,7 +882,7 @@ class StreamSource(OnlineSource):
             kwargs["worker_allocation_mode"] = "static"
 
         function.add_v3io_stream_trigger(
-            path,
+            url,
             self.name,
             self.attributes["group"],
             self.attributes["seek_to"],

mlrun/datastore/targets.py

@@ -524,7 +524,12 @@ class BaseStoreTarget(DataTargetBase):
             store, path_in_store, target_path = self._get_store_and_path()
             target_path = generate_path_with_chunk(self, chunk_id, target_path)
             file_system = store.filesystem
-            if file_system.protocol == "file":
+            if (
+                file_system.protocol == "file"
+                # fsspec 2023.10.0 changed protocol from "file" to ("file", "local")
+                or isinstance(file_system.protocol, (tuple, list))
+                and "file" in file_system.protocol
+            ):
                 dir = os.path.dirname(target_path)
                 if dir:
                     os.makedirs(dir, exist_ok=True)