mlrun 1.7.0rc14__py3-none-any.whl → 1.7.0rc22__py3-none-any.whl

mlrun/runtimes/utils.py

@@ -20,17 +20,17 @@ from io import StringIO
 from sys import stderr
 
 import pandas as pd
-from kubernetes import client
 
 import mlrun
 import mlrun.common.constants
+import mlrun.common.constants as mlrun_constants
 import mlrun.common.schemas
 import mlrun.utils.regex
 from mlrun.artifacts import TableArtifact
+from mlrun.common.runtimes.constants import RunLabels
 from mlrun.config import config
 from mlrun.errors import err_to_str
 from mlrun.frameworks.parallel_coordinates import gen_pcp_plot
-from mlrun.runtimes.constants import RunLabels
 from mlrun.runtimes.generators import selector
 from mlrun.utils import get_in, helpers, logger, verify_field_regex
 
@@ -39,9 +39,6 @@ class RunError(Exception):
     pass
 
 
-mlrun_key = "mlrun/"
-
-
 class _ContextStore:
     def __init__(self):
         self._context = None
@@ -280,43 +277,6 @@ def get_item_name(item, attr="name"):
         return getattr(item, attr, None)
 
 
-def apply_kfp(modify, cop, runtime):
-    modify(cop)
-
-    # Have to do it here to avoid circular dependencies
-    from .pod import AutoMountType
-
-    if AutoMountType.is_auto_modifier(modify):
-        runtime.spec.disable_auto_mount = True
-
-    api = client.ApiClient()
-    for k, v in cop.pod_labels.items():
-        runtime.metadata.labels[k] = v
-    for k, v in cop.pod_annotations.items():
-        runtime.metadata.annotations[k] = v
-    if cop.container.env:
-        env_names = [
-            e.name if hasattr(e, "name") else e["name"] for e in runtime.spec.env
-        ]
-        for e in api.sanitize_for_serialization(cop.container.env):
-            name = e["name"]
-            if name in env_names:
-                runtime.spec.env[env_names.index(name)] = e
-            else:
-                runtime.spec.env.append(e)
-                env_names.append(name)
-        cop.container.env.clear()
-
-    if cop.volumes and cop.container.volume_mounts:
-        vols = api.sanitize_for_serialization(cop.volumes)
-        mounts = api.sanitize_for_serialization(cop.container.volume_mounts)
-        runtime.spec.update_vols_and_mounts(vols, mounts)
-        cop.volumes.clear()
-        cop.container.volume_mounts.clear()
-
-    return runtime
-
-
 def verify_limits(
     resources_field_name,
     mem=None,
@@ -410,10 +370,10 @@ def generate_resources(mem=None, cpu=None, gpus=None, gpu_type="nvidia.com/gpu")
 
 
 def get_func_selector(project, name=None, tag=None):
-    s = [f"{mlrun_key}project={project}"]
+    s = [f"{mlrun_constants.MLRunInternalLabels.project}={project}"]
     if name:
-        s.append(f"{mlrun_key}function={name}")
-        s.append(f"{mlrun_key}tag={tag or 'latest'}")
+        s.append(f"{mlrun_constants.MLRunInternalLabels.function}={name}")
+        s.append(f"{mlrun_constants.MLRunInternalLabels.tag}={tag or 'latest'}")
     return s
 
 
@@ -476,6 +436,7 @@ def enrich_run_labels(
 ):
     labels_enrichment = {
         RunLabels.owner: os.environ.get("V3IO_USERNAME") or getpass.getuser(),
+        # TODO: remove this in 1.9.0
         RunLabels.v3io_user: os.environ.get("V3IO_USERNAME"),
     }
     labels_to_enrich = labels_to_enrich or RunLabels.all()

mlrun/serving/__init__.py

@@ -22,10 +22,17 @@ __all__ = [
     "RouterStep",
     "QueueStep",
     "ErrorStep",
+    "MonitoringApplicationStep",
 ]
 
 from .routers import ModelRouter, VotingEnsemble  # noqa
 from .server import GraphContext, GraphServer, create_graph_server  # noqa
-from .states import ErrorStep, QueueStep, RouterStep, TaskStep  # noqa
+from .states import (
+    ErrorStep,
+    QueueStep,
+    RouterStep,
+    TaskStep,
+    MonitoringApplicationStep,
+)  # noqa
 from .v1_serving import MLModelServer, new_v1_model_server  # noqa
 from .v2_serving import V2ModelServer  # noqa

mlrun/serving/server.py

@@ -387,7 +387,7 @@ def v2_serving_handler(context, event, get_body=False):
 
 
 def create_graph_server(
-    parameters={},
+    parameters=None,
     load_mode=None,
     graph=None,
     verbose=False,
@@ -403,6 +403,7 @@ def create_graph_server(
         server.graph.add_route("my", class_name=MyModelClass, model_path="{path}", z=100)
         print(server.test("/v2/models/my/infer", testdata))
     """
+    parameters = parameters or {}
     server = GraphServer(graph, parameters, load_mode, verbose=verbose, **kwargs)
     server.set_current_function(
         current_function or os.environ.get("SERVING_CURRENT_FUNCTION", "")

mlrun/serving/states.py

@@ -12,7 +12,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__all__ = ["TaskStep", "RouterStep", "RootFlowStep", "ErrorStep"]
+__all__ = [
+    "TaskStep",
+    "RouterStep",
+    "RootFlowStep",
+    "ErrorStep",
+    "MonitoringApplicationStep",
+]
 
 import os
 import pathlib
@@ -55,6 +61,7 @@ class StepKinds:
     choice = "choice"
     root = "root"
     error_step = "error_step"
+    monitoring_application = "monitoring_application"
 
 
 _task_step_fields = [
@@ -485,13 +492,15 @@ class TaskStep(BaseStep):
                 class_args[key] = arg
         class_args.update(extra_kwargs)
 
-        # add common args (name, context, ..) only if target class can accept them
-        argspec = getfullargspec(class_object)
-        for key in ["name", "context", "input_path", "result_path", "full_event"]:
-            if argspec.varkw or key in argspec.args:
-                class_args[key] = getattr(self, key)
-        if argspec.varkw or "graph_step" in argspec.args:
-            class_args["graph_step"] = self
+        if not isinstance(self, MonitoringApplicationStep):
+            # add common args (name, context, ..) only if target class can accept them
+            argspec = getfullargspec(class_object)
+
+            for key in ["name", "context", "input_path", "result_path", "full_event"]:
+                if argspec.varkw or key in argspec.args:
+                    class_args[key] = getattr(self, key)
+            if argspec.varkw or "graph_step" in argspec.args:
+                class_args["graph_step"] = self
         return class_args
 
     def get_step_class_object(self, namespace):
@@ -582,6 +591,39 @@ class TaskStep(BaseStep):
         return event
 
 
+class MonitoringApplicationStep(TaskStep):
+    """monitoring application execution step, runs users class code"""
+
+    kind = "monitoring_application"
+    _default_class = ""
+
+    def __init__(
+        self,
+        class_name: Union[str, type] = None,
+        class_args: dict = None,
+        handler: str = None,
+        name: str = None,
+        after: list = None,
+        full_event: bool = None,
+        function: str = None,
+        responder: bool = None,
+        input_path: str = None,
+        result_path: str = None,
+    ):
+        super().__init__(
+            class_name=class_name,
+            class_args=class_args,
+            handler=handler,
+            name=name,
+            after=after,
+            full_event=full_event,
+            function=function,
+            responder=responder,
+            input_path=input_path,
+            result_path=result_path,
+        )
+
+
 class ErrorStep(TaskStep):
     """error execution step, runs a class or handler"""
 
@@ -1323,6 +1365,7 @@ classes_map = {
     "flow": FlowStep,
     "queue": QueueStep,
     "error_step": ErrorStep,
+    "monitoring_application": MonitoringApplicationStep,
 }
 
 

mlrun/serving/utils.py

@@ -46,6 +46,15 @@ def _update_result_body(result_path, event_body, result):
 class StepToDict:
     """auto serialization of graph steps to a python dictionary"""
 
+    meta_keys = [
+        "context",
+        "name",
+        "input_path",
+        "result_path",
+        "full_event",
+        "kwargs",
+    ]
+
     def to_dict(self, fields: list = None, exclude: list = None, strip: bool = False):
         """convert the step object to a python dictionary"""
         fields = fields or getattr(self, "_dict_fields", None)
@@ -54,24 +63,16 @@ class StepToDict:
         if exclude:
             fields = [field for field in fields if field not in exclude]
 
-        meta_keys = [
-            "context",
-            "name",
-            "input_path",
-            "result_path",
-            "full_event",
-            "kwargs",
-        ]
         args = {
             key: getattr(self, key)
             for key in fields
-            if getattr(self, key, None) is not None and key not in meta_keys
+            if getattr(self, key, None) is not None and key not in self.meta_keys
         }
         # add storey kwargs or extra kwargs
         if "kwargs" in fields and (hasattr(self, "kwargs") or hasattr(self, "_kwargs")):
             kwargs = getattr(self, "kwargs", {}) or getattr(self, "_kwargs", {})
             for key, value in kwargs.items():
-                if key not in meta_keys:
+                if key not in self.meta_keys:
                     args[key] = value
 
         mod_name = self.__class__.__module__
@@ -80,7 +81,9 @@ class StepToDict:
             class_path = f"{mod_name}.{class_path}"
         struct = {
             "class_name": class_path,
-            "name": self.name or self.__class__.__name__,
+            "name": self.name
+            if hasattr(self, "name") and self.name
+            else self.__class__.__name__,
             "class_args": args,
         }
         if hasattr(self, "_STEP_KIND"):
@@ -94,6 +97,11 @@ class StepToDict:
         return struct
 
 
+class MonitoringApplicationToDict(StepToDict):
+    _STEP_KIND = "monitoring_application"
+    meta_keys = []
+
+
 class RouterToDict(StepToDict):
     _STEP_KIND = "router"
 

mlrun/serving/v2_serving.py

@@ -528,7 +528,11 @@ def _init_endpoint_record(
         return None
 
     # Generating version model value based on the model name and model version
-    if model.version:
+    if model.model_path and model.model_path.startswith("store://"):
+        # Enrich the model server with the model artifact metadata
+        model.get_model()
+        model.version = model.model_spec.tag
+        model.labels = model.model_spec.labels
         versioned_model_name = f"{model.name}:{model.version}"
     else:
         versioned_model_name = f"{model.name}:latest"

mlrun/track/tracker.py

@@ -31,8 +31,9 @@ class Tracker(ABC):
     * Offline: Manually importing models and artifacts into an MLRun project using the `import_x` methods.
     """
 
+    @staticmethod
     @abstractmethod
-    def is_enabled(self) -> bool:
+    def is_enabled() -> bool:
         """
         Checks if tracker is enabled.
 

mlrun/utils/async_http.py

@@ -24,7 +24,7 @@ from aiohttp_retry import ExponentialRetry, RequestParams, RetryClient, RetryOpt
 from aiohttp_retry.client import _RequestContext
 
 from mlrun.config import config
-from mlrun.errors import err_to_str
+from mlrun.errors import err_to_str, raise_for_status
 
 from .helpers import logger as mlrun_logger
 
@@ -46,12 +46,21 @@ class AsyncClientWithRetry(RetryClient):
         *args,
         **kwargs,
     ):
+        # do not retry on PUT / PATCH as they might have side effects (not truly idempotent)
+        blacklisted_methods = (
+            blacklisted_methods
+            if blacklisted_methods is not None
+            else [
+                "POST",
+                "PUT",
+                "PATCH",
+            ]
+        )
         super().__init__(
             *args,
             retry_options=ExponentialRetryOverride(
                 retry_on_exception=retry_on_exception,
-                # do not retry on PUT / PATCH as they might have side effects (not truly idempotent)
-                blacklisted_methods=blacklisted_methods or ["POST", "PUT", "PATCH"],
+                blacklisted_methods=blacklisted_methods,
                 attempts=max_retries,
                 statuses=retry_on_status_codes,
                 factor=retry_backoff_factor,
@@ -63,6 +72,12 @@ class AsyncClientWithRetry(RetryClient):
             **kwargs,
         )
 
+    def methods_blacklist_update_required(self, new_blacklist: str):
+        self._retry_options: ExponentialRetryOverride
+        return set(self._retry_options.blacklisted_methods).difference(
+            set(new_blacklist)
+        )
+
     def _make_requests(
         self,
         params_list: list[RequestParams],
@@ -173,7 +188,7 @@ class _CustomRequestContext(_RequestContext):
                 last_attempt = current_attempt == self._retry_options.attempts
                 if self._is_status_code_ok(response.status) or last_attempt:
                     if self._raise_for_status:
-                        response.raise_for_status()
+                        raise_for_status(response)
 
                     self._response = response
                     return response
@@ -275,6 +290,11 @@ class _CustomRequestContext(_RequestContext):
                 if isinstance(exc.os_error, exc_type):
                     return
         if exc.__cause__:
-            return self.verify_exception_type(exc.__cause__)
+            # If the cause exception is retriable, return, otherwise, raise the original exception
+            try:
+                self.verify_exception_type(exc.__cause__)
+            except Exception:
+                raise exc
+            return
         else:
             raise exc