mlrun 1.7.0rc4__py3-none-any.whl → 1.7.0rc20__py3-none-any.whl

mlrun/serving/states.py

@@ -14,19 +14,21 @@
 
 __all__ = ["TaskStep", "RouterStep", "RootFlowStep", "ErrorStep"]
 
-import asyncio
 import os
 import pathlib
 import traceback
 from copy import copy, deepcopy
 from inspect import getfullargspec, signature
-from typing import Union
+from typing import Any, Union
 
 import mlrun
 
 from ..config import config
 from ..datastore import get_stream_pusher
-from ..datastore.utils import parse_kafka_url
+from ..datastore.utils import (
+    get_kafka_brokers_from_dict,
+    parse_kafka_url,
+)
 from ..errors import MLRunInvalidArgumentError, err_to_str
 from ..model import ModelObj, ObjectDict
 from ..platforms.iguazio import parse_path
@@ -325,7 +327,7 @@ class BaseStep(ModelObj):
             parent = self._parent
         else:
             raise GraphError(
-                f"step {self.name} parent is not set or its not part of a graph"
+                f"step {self.name} parent is not set or it's not part of a graph"
             )
 
         name, step = params_to_step(
@@ -347,6 +349,36 @@ class BaseStep(ModelObj):
         parent._last_added = step
         return step
 
+    def set_flow(
+        self,
+        steps: list[Union[str, StepToDict, dict[str, Any]]],
+        force: bool = False,
+    ):
+        """set list of steps as downstream from this step, in the order specified. This will overwrite any existing
+        downstream steps.
+
+        :param steps: list of steps to follow this one
+        :param force: whether to overwrite existing downstream steps. If False, this method will fail if any downstream
+        steps have already been defined. Defaults to False.
+        :return: the last step added to the flow
+
+        example:
+            The below code sets the downstream nodes of step1 by using a list of steps (provided to `set_flow()`) and a
+            single step (provided to `to()`), resulting in the graph (step1 -> step2 -> step3 -> step4).
+            Notice that using `force=True` is required in case step1 already had downstream nodes (e.g. if the existing
+            graph is step1 -> step2_old) and that following the execution of this code the existing downstream steps
+            are removed. If the intention is to split the graph (and not to overwrite), please use `to()`.
+
+            step1.set_flow(
+                [
+                    dict(name="step2", handler="step2_handler"),
+                    dict(name="step3", class_name="Step3Class"),
+                ],
+                force=True,
+            ).to(dict(name="step4", class_name="Step4Class"))
+        """
+        raise NotImplementedError("set_flow() can only be called on a FlowStep")
+
 
 class TaskStep(BaseStep):
     """task execution step, runs a class or handler"""
@@ -1161,19 +1193,11 @@ class FlowStep(BaseStep):
         if self._controller:
             # async flow (using storey)
             event._awaitable_result = None
-            if self.context.is_mock:
-                resp = self._controller.emit(
-                    event, return_awaitable_result=self._wait_for_result
-                )
-                if self._wait_for_result and resp:
-                    return resp.await_result()
-            else:
-                resp_awaitable = self._controller.emit(
-                    event, await_result=self._wait_for_result
-                )
-                if self._wait_for_result:
-                    return resp_awaitable
-                return self._await_and_return_id(resp_awaitable, event)
+            resp = self._controller.emit(
+                event, return_awaitable_result=self._wait_for_result
+            )
+            if self._wait_for_result and resp:
+                return resp.await_result()
             event = copy(event)
             event.body = {"id": event.id}
             return event
@@ -1213,18 +1237,9 @@ class FlowStep(BaseStep):
         """wait for completion of run in async flows"""
 
         if self._controller:
-            if asyncio.iscoroutinefunction(self._controller.await_termination):
-
-                async def terminate_and_await_termination():
-                    if hasattr(self._controller, "terminate"):
-                        await self._controller.terminate()
-                    return await self._controller.await_termination()
-
-                return terminate_and_await_termination()
-            else:
-                if hasattr(self._controller, "terminate"):
-                    self._controller.terminate()
-                return self._controller.await_termination()
+            if hasattr(self._controller, "terminate"):
+                self._controller.terminate()
+            return self._controller.await_termination()
 
     def plot(self, filename=None, format=None, source=None, targets=None, **kw):
         """plot/save graph using graphviz
@@ -1273,6 +1288,27 @@ class FlowStep(BaseStep):
                 )
             self[step_name].after_step(name)
 
+    def set_flow(
+        self,
+        steps: list[Union[str, StepToDict, dict[str, Any]]],
+        force: bool = False,
+    ):
+        if not force and self.steps:
+            raise mlrun.errors.MLRunInvalidArgumentError(
+                "set_flow() called on a step that already has downstream steps. "
+                "If you want to overwrite existing steps, set force=True."
+            )
+
+        self.steps = None
+        step = self
+        for next_step in steps:
+            if isinstance(next_step, dict):
+                step = step.to(**next_step)
+            else:
+                step = step.to(next_step)
+
+        return step
+
 
 class RootFlowStep(FlowStep):
     """root flow step"""
@@ -1512,13 +1548,11 @@ def _init_async_objects(context, steps):
                     endpoint = None
                     options = {}
                     options.update(step.options)
-                    kafka_bootstrap_servers = options.pop(
-                        "kafka_bootstrap_servers", None
-                    )
-                    if stream_path.startswith("kafka://") or kafka_bootstrap_servers:
-                        topic, bootstrap_servers = parse_kafka_url(
-                            stream_path, kafka_bootstrap_servers
-                        )
+
+                    kafka_brokers = get_kafka_brokers_from_dict(options, pop=True)
+
+                    if stream_path.startswith("kafka://") or kafka_brokers:
+                        topic, brokers = parse_kafka_url(stream_path, kafka_brokers)
 
                         kafka_producer_options = options.pop(
                             "kafka_producer_options", None
@@ -1526,7 +1560,7 @@ def _init_async_objects(context, steps):
 
                         step._async_object = storey.KafkaTarget(
                             topic=topic,
-                            bootstrap_servers=bootstrap_servers,
+                            brokers=brokers,
                             producer_options=kafka_producer_options,
                             context=context,
                             **options,
@@ -1568,12 +1602,8 @@ def _init_async_objects(context, steps):
     source_args = context.get_param("source_args", {})
     explicit_ack = is_explicit_ack_supported(context) and mlrun.mlconf.is_explicit_ack()
 
-    if context.is_mock:
-        source_class = storey.SyncEmitSource
-    else:
-        source_class = storey.AsyncEmitSource
-
-    default_source = source_class(
+    # TODO: Change to AsyncEmitSource once we can drop support for nuclio<1.12.10
+    default_source = storey.SyncEmitSource(
         context=context,
         explicit_ack=explicit_ack,
         **source_args,

mlrun/serving/v2_serving.py

@@ -21,6 +21,7 @@ import mlrun.common.model_monitoring
 import mlrun.common.schemas.model_monitoring
 from mlrun.artifacts import ModelArtifact  # noqa: F401
 from mlrun.config import config
+from mlrun.errors import err_to_str
 from mlrun.utils import logger, now_date
 
 from ..common.helpers import parse_versioned_object_uri
@@ -62,11 +63,11 @@ class V2ModelServer(StepToDict):
             class MyClass(V2ModelServer):
                 def load(self):
                     # load and initialize the model and/or other elements
-                    model_file, extra_data = self.get_model(suffix='.pkl')
+                    model_file, extra_data = self.get_model(suffix=".pkl")
                     self.model = load(open(model_file, "rb"))
 
                 def predict(self, request):
-                    events = np.array(request['inputs'])
+                    events = np.array(request["inputs"])
                     dmatrix = xgb.DMatrix(events)
                     result: xgb.DMatrix = self.model.predict(dmatrix)
                     return {"outputs": result.tolist()}
@@ -175,9 +176,9 @@ class V2ModelServer(StepToDict):
         ::
 
             def load(self):
-                model_file, extra_data = self.get_model(suffix='.pkl')
+                model_file, extra_data = self.get_model(suffix=".pkl")
                 self.model = load(open(model_file, "rb"))
-                categories = extra_data['categories'].as_df()
+                categories = extra_data["categories"].as_df()
 
         Parameters
         ----------
@@ -523,7 +524,7 @@ def _init_endpoint_record(
             graph_server.function_uri
         )
     except Exception as e:
-        logger.error("Failed to parse function URI", exc=e)
+        logger.error("Failed to parse function URI", exc=err_to_str(e))
         return None
 
     # Generating version model value based on the model name and model version
@@ -576,9 +577,9 @@ def _init_endpoint_record(
             )
 
         except Exception as e:
-            logger.error("Failed to create endpoint record", exc=e)
+            logger.error("Failed to create endpoint record", exc=err_to_str(e))
 
     except Exception as e:
-        logger.error("Failed to retrieve model endpoint object", exc=e)
+        logger.error("Failed to retrieve model endpoint object", exc=err_to_str(e))
 
     return uid

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

mlrun/utils/helpers.py

@@ -39,7 +39,7 @@ import pandas
 import semver
 import yaml
 from dateutil import parser
-from deprecated import deprecated
+from mlrun_pipelines.models import PipelineRun
 from pandas._libs.tslibs.timestamps import Timedelta, Timestamp
 from yaml.representer import RepresenterError
 
@@ -76,19 +76,6 @@ class OverwriteBuildParamsWarning(FutureWarning):
     pass
 
 
-# TODO: remove in 1.7.0
-@deprecated(
-    version="1.5.0",
-    reason="'parse_versioned_object_uri' will be removed from this file in 1.7.0, use "
-    "'mlrun.common.helpers.parse_versioned_object_uri' instead",
-    category=FutureWarning,
-)
-def parse_versioned_object_uri(uri: str, default_project: str = ""):
-    return mlrun.common.helpers.parse_versioned_object_uri(
-        uri=uri, default_project=default_project
-    )
-
-
 class StorePrefix:
     """map mlrun store objects to prefixes"""
 
@@ -119,14 +106,9 @@ class StorePrefix:
 
 
 def get_artifact_target(item: dict, project=None):
-    if is_legacy_artifact(item):
-        db_key = item.get("db_key")
-        project_str = project or item.get("project")
-        tree = item.get("tree")
-    else:
-        db_key = item["spec"].get("db_key")
-        project_str = project or item["metadata"].get("project")
-        tree = item["metadata"].get("tree")
+    db_key = item["spec"].get("db_key")
+    project_str = project or item["metadata"].get("project")
+    tree = item["metadata"].get("tree")
 
     kind = item.get("kind")
     if kind in ["dataset", "model", "artifact"] and db_key:
@@ -135,11 +117,15 @@ def get_artifact_target(item: dict, project=None):
             target = f"{target}@{tree}"
         return target
 
-    return (
-        item.get("target_path")
-        if is_legacy_artifact(item)
-        else item["spec"].get("target_path")
-    )
+    return item["spec"].get("target_path")
+
+
+# TODO: left for migrations testing purposes. Remove in 1.8.0.
+def is_legacy_artifact(artifact):
+    if isinstance(artifact, dict):
+        return "metadata" not in artifact
+    else:
+        return not hasattr(artifact, "metadata")
 
 
 logger = create_logger(config.log_level, config.log_formatter, "mlrun", sys.stdout)
@@ -195,8 +181,12 @@ def verify_field_regex(
             )
             if mode == mlrun.common.schemas.RegexMatchModes.all:
                 if raise_on_failure:
+                    if len(field_name) > max_chars:
+                        field_name = field_name[:max_chars] + "...truncated"
+                    if len(field_value) > max_chars:
+                        field_value = field_value[:max_chars] + "...truncated"
                     raise mlrun.errors.MLRunInvalidArgumentError(
-                        f"Field '{field_name[:max_chars]}' is malformed. '{field_value[:max_chars]}' "
+                        f"Field '{field_name}' is malformed. '{field_value}' "
                         f"does not match required pattern: {pattern}"
                     )
                 return False
@@ -437,7 +427,7 @@ class LogBatchWriter:
 
 def get_in(obj, keys, default=None):
     """
-    >>> get_in({'a': {'b': 1}}, 'a.b')
+    >>> get_in({"a": {"b": 1}}, "a.b")
     1
     """
     if isinstance(keys, str):
@@ -801,34 +791,6 @@ def gen_html_table(header, rows=None):
     return style + '<table class="tg">\n' + out + "</table>\n\n"
 
 
-def new_pipe_metadata(
-    artifact_path: str = None,
-    cleanup_ttl: int = None,
-    op_transformers: list[typing.Callable] = None,
-):
-    from kfp.dsl import PipelineConf
-
-    def _set_artifact_path(task):
-        from kubernetes import client as k8s_client
-
-        task.add_env_variable(
-            k8s_client.V1EnvVar(name="MLRUN_ARTIFACT_PATH", value=artifact_path)
-        )
-        return task
-
-    conf = PipelineConf()
-    cleanup_ttl = cleanup_ttl or int(config.kfp_ttl)
-
-    if cleanup_ttl:
-        conf.set_ttl_seconds_after_finished(cleanup_ttl)
-    if artifact_path:
-        conf.add_op_transformer(_set_artifact_path)
-    if op_transformers:
-        for op_transformer in op_transformers:
-            conf.add_op_transformer(op_transformer)
-    return conf
-
-
 def _convert_python_package_version_to_image_tag(version: typing.Optional[str]):
     return (
         version.replace("+", "-").replace("0.0.0-", "") if version is not None else None
@@ -1015,17 +977,27 @@ def get_ui_url(project, uid=None):
     return url
 
 
+def get_model_endpoint_url(project, model_name, model_endpoint_id):
+    url = ""
+    if mlrun.mlconf.resolve_ui_url():
+        url = f"{mlrun.mlconf.resolve_ui_url()}/{mlrun.mlconf.ui.projects_prefix}/{project}/models"
+        if model_name:
+            url += f"/model-endpoints/{model_name}/{model_endpoint_id}/overview"
+    return url
+
+
 def get_workflow_url(project, id=None):
     url = ""
     if mlrun.mlconf.resolve_ui_url():
-        url = "{}/{}/{}/jobs/monitor-workflows/workflow/{}".format(
-            mlrun.mlconf.resolve_ui_url(), mlrun.mlconf.ui.projects_prefix, project, id
+        url = (
+            f"{mlrun.mlconf.resolve_ui_url()}/{mlrun.mlconf.ui.projects_prefix}"
+            f"/{project}/jobs/monitor-workflows/workflow/{id}"
         )
     return url
 
 
 def are_strings_in_exception_chain_messages(
-    exception: Exception, strings_list=list[str]
+    exception: Exception, strings_list: list[str]
 ) -> bool:
     while exception is not None:
         if any([string in str(exception) for string in strings_list]):
@@ -1138,7 +1110,7 @@ def get_function(function, namespace):
 
 
 def get_handler_extended(
-    handler_path: str, context=None, class_args: dict = {}, namespaces=None
+    handler_path: str, context=None, class_args: dict = None, namespaces=None
 ):
     """get function handler from [class_name::]handler string
 
@@ -1148,6 +1120,7 @@ def get_handler_extended(
     :param namespaces:    one or list of namespaces/modules to search the handler in
     :return: function handler (callable)
     """
+    class_args = class_args or {}
     if "::" not in handler_path:
         return get_function(handler_path, namespaces)
 
@@ -1224,7 +1197,7 @@ def calculate_dataframe_hash(dataframe: pandas.DataFrame):
     return hashlib.sha1(pandas.util.hash_pandas_object(dataframe).values).hexdigest()
 
 
-def template_artifact_path(artifact_path, project, run_uid="project"):
+def template_artifact_path(artifact_path, project, run_uid=None):
     """
     Replace {{run.uid}} with the run uid and {{project}} with the project name in the artifact path.
     If no run uid is provided, the word `project` will be used instead as it is assumed to be a project
@@ -1232,6 +1205,7 @@ def template_artifact_path(artifact_path, project, run_uid="project"):
     """
     if not artifact_path:
         return artifact_path
+    run_uid = run_uid or "project"
     artifact_path = artifact_path.replace("{{run.uid}}", run_uid)
     artifact_path = _fill_project_path_template(artifact_path, project)
     return artifact_path
@@ -1291,13 +1265,6 @@ def str_to_timestamp(time_str: str, now_time: Timestamp = None):
     return Timestamp(time_str)
 
 
-def is_legacy_artifact(artifact):
-    if isinstance(artifact, dict):
-        return "metadata" not in artifact
-    else:
-        return not hasattr(artifact, "metadata")
-
-
 def is_link_artifact(artifact):
     if isinstance(artifact, dict):
         return (
@@ -1307,7 +1274,7 @@ def is_link_artifact(artifact):
         return artifact.kind == mlrun.common.schemas.ArtifactCategories.link.value
 
 
-def format_run(run: dict, with_project=False) -> dict:
+def format_run(run: PipelineRun, with_project=False) -> dict:
     fields = [
         "id",
         "name",
@@ -1344,17 +1311,17 @@ def format_run(run: dict, with_project=False) -> dict:
     # pipelines are yet to populate the status or workflow has failed
     # as observed https://jira.iguazeng.com/browse/ML-5195
     # set to unknown to ensure a status is returned
-    if run["status"] is None:
-        run["status"] = inflection.titleize(mlrun.runtimes.constants.RunStates.unknown)
+    if run.get("status", None) is None:
+        run["status"] = inflection.titleize(
+            mlrun.common.runtimes.constants.RunStates.unknown
+        )
 
     return run
 
 
 def get_in_artifact(artifact: dict, key, default=None, raise_on_missing=False):
     """artifact can be dict or Artifact object"""
-    if is_legacy_artifact(artifact):
-        return artifact.get(key, default)
-    elif key == "kind":
+    if key == "kind":
         return artifact.get(key, default)
     else:
         for block in ["metadata", "spec", "status"]:
@@ -1405,6 +1372,18 @@ def as_number(field_name, field_value):
 
 
 def filter_warnings(action, category):
+    """
+    Decorator to filter warnings
+
+    Example::
+        @filter_warnings("ignore", FutureWarning)
+        def my_function():
+            pass
+
+    :param action:      one of "error", "ignore", "always", "default", "module", or "once"
+    :param category:    a class that the warning must be a subclass of
+    """
+
     def decorator(function):
         def wrapper(*args, **kwargs):
             # context manager that copies and, upon exit, restores the warnings filter and the showwarning() function.
@@ -1562,3 +1541,90 @@ def is_safe_path(base, filepath, is_symlink=False):
         os.path.abspath(filepath) if not is_symlink else os.path.realpath(filepath)
     )
     return base == os.path.commonpath((base, resolved_filepath))
+
+
+def get_serving_spec():
+    data = None
+
+    # we will have the serving spec in either mounted config map
+    # or env depending on the size of the spec and configuration
+
+    try:
+        with open(mlrun.common.constants.MLRUN_SERVING_SPEC_PATH) as f:
+            data = f.read()
+    except FileNotFoundError:
+        pass
+
+    if data is None:
+        data = os.environ.get("SERVING_SPEC_ENV", "")
+        if not data:
+            raise mlrun.errors.MLRunInvalidArgumentError(
+                "Failed to find serving spec in env var or config file"
+            )
+    spec = json.loads(data)
+    return spec
+
+
+def additional_filters_warning(additional_filters, class_name):
+    if additional_filters and any(additional_filters):
+        mlrun.utils.logger.warn(
+            f"additional_filters parameter is not supported in {class_name},"
+            f" parameter has been ignored."
+        )
+
+
+def validate_component_version_compatibility(
+    component_name: typing.Literal["iguazio", "nuclio"], *min_versions: str
+):
+    """
+    :param component_name: Name of the component to validate compatibility for.
+    :param min_versions: Valid minimum version(s) required, assuming no 2 versions has equal major and minor.
+    """
+    parsed_min_versions = [
+        semver.VersionInfo.parse(min_version) for min_version in min_versions
+    ]
+    parsed_current_version = None
+    component_current_version = None
+    try:
+        if component_name == "iguazio":
+            component_current_version = mlrun.mlconf.igz_version
+            parsed_current_version = mlrun.mlconf.get_parsed_igz_version()
+
+            if parsed_current_version:
+                # ignore pre-release and build metadata, as iguazio version always has them, and we only care about the
+                # major, minor, and patch versions
+                parsed_current_version = semver.VersionInfo.parse(
+                    f"{parsed_current_version.major}.{parsed_current_version.minor}.{parsed_current_version.patch}"
+                )
+        if component_name == "nuclio":
+            component_current_version = mlrun.mlconf.nuclio_version
+            parsed_current_version = semver.VersionInfo.parse(
+                mlrun.mlconf.nuclio_version
+            )
+        if not parsed_current_version:
+            return True
+    except ValueError:
+        # only log when version is set but invalid
+        if component_current_version:
+            logger.warning(
+                "Unable to parse current version, assuming compatibility",
+                component_name=component_name,
+                current_version=component_current_version,
+                min_versions=min_versions,
+            )
+        return True
+
+    parsed_min_versions.sort(reverse=True)
+    for parsed_min_version in parsed_min_versions:
+        if parsed_current_version < parsed_min_version:
+            return False
+    return True
+
+
+def format_alert_summary(
+    alert: mlrun.common.schemas.AlertConfig, event_data: mlrun.common.schemas.Event
+) -> str:
+    result = alert.summary.replace("{{project}}", alert.project)
+    result = result.replace("{{name}}", alert.name)
+    result = result.replace("{{entity}}", event_data.entity.ids[0])
+    return result

mlrun/utils/http.py

@@ -122,7 +122,7 @@ class HTTPSessionWithRetry(requests.Session):
 
                 self._logger.warning(
                     "Error during request handling, retrying",
-                    exc=str(exc),
+                    exc=err_to_str(exc),
                     retry_count=retry_count,
                     url=url,
                     method=method,