Adding option to control retry for any aws source

Adding logs to make it transparent what is going on in SageMaker

Author: treff7es

metadata-ingestion/src/datahub/ingestion/source/aws/aws_common.py

@@ -107,6 +107,14 @@ class AwsConnectionConfig(ConfigModel):
         default=None,
         description="A set of proxy configs to use with AWS. See the [botocore.config](https://botocore.amazonaws.com/v1/documentation/api/latest/reference/config.html) docs for details.",
     )
+    aws_retry_num: int = Field(
+        default=5,
+        description="Number of times to retry failed AWS requests. See the [botocore.retry](https://boto3.amazonaws.com/v1/documentation/api/latest/guide/retries.html) docs for details.",
+    )
+    aws_retry_mode: Literal["legacy", "standard", "adaptive"] = Field(
+        default="standard",
+        description="Retry mode to use for failed AWS requests. See the [botocore.retry](https://boto3.amazonaws.com/v1/documentation/api/latest/guide/retries.html) docs for details.",
+    )
 
     read_timeout: float = Field(
         default=DEFAULT_TIMEOUT,
@@ -199,6 +207,10 @@ def _aws_config(self) -> Config:
         return Config(
             proxies=self.aws_proxy,
             read_timeout=self.read_timeout,
+            retries={
+                "max_attempts": self.aws_retry_num,
+                "mode": self.aws_retry_mode,
+            },
             **self.aws_advanced_config,
         )
 

metadata-ingestion/src/datahub/ingestion/source/aws/sagemaker.py

@@ -1,3 +1,4 @@
+import logging
 from collections import defaultdict
 from typing import TYPE_CHECKING, DefaultDict, Dict, Iterable, List, Optional
 
@@ -36,6 +37,8 @@
 if TYPE_CHECKING:
     from mypy_boto3_sagemaker import SageMakerClient
 
+logger = logging.getLogger(__name__)
+
 
 @platform_name("SageMaker")
 @config_class(SagemakerSourceConfig)
@@ -75,6 +78,7 @@ def get_workunit_processors(self) -> List[Optional[MetadataWorkUnitProcessor]]:
         ]
 
     def get_workunits_internal(self) -> Iterable[MetadataWorkUnit]:
+        logger.info("Starting SageMaker ingestion...")
         # get common lineage graph
         lineage_processor = LineageProcessor(
             sagemaker_client=self.sagemaker_client, env=self.env, report=self.report
@@ -83,6 +87,7 @@ def get_workunits_internal(self) -> Iterable[MetadataWorkUnit]:
 
         # extract feature groups if specified
         if self.source_config.extract_feature_groups:
+            logger.info("Extracting feature groups...")
             feature_group_processor = FeatureGroupProcessor(
                 sagemaker_client=self.sagemaker_client, env=self.env, report=self.report
             )
@@ -95,6 +100,7 @@ def get_workunits_internal(self) -> Iterable[MetadataWorkUnit]:
 
         # extract jobs if specified
         if self.source_config.extract_jobs is not False:
+            logger.info("Extracting jobs...")
             job_processor = JobProcessor(
                 sagemaker_client=self.client_factory.get_client,
                 env=self.env,
@@ -109,6 +115,8 @@ def get_workunits_internal(self) -> Iterable[MetadataWorkUnit]:
 
         # extract models if specified
         if self.source_config.extract_models:
+            logger.info("Extracting models...")
+
             model_processor = ModelProcessor(
                 sagemaker_client=self.sagemaker_client,
                 env=self.env,

metadata-ingestion/src/datahub/ingestion/source/aws/sagemaker_processors/common.py

@@ -40,8 +40,11 @@ class SagemakerSourceReport(StaleEntityRemovalSourceReport):
     groups_scanned = 0
     models_scanned = 0
     jobs_scanned = 0
+    jobs_processed = 0
     datasets_scanned = 0
     filtered: List[str] = field(default_factory=list)
+    model_endpoint_lineage = 0
+    model_group_lineage = 0
 
     def report_feature_group_scanned(self) -> None:
         self.feature_groups_scanned += 1
@@ -58,6 +61,9 @@ def report_group_scanned(self) -> None:
     def report_model_scanned(self) -> None:
         self.models_scanned += 1
 
+    def report_job_processed(self) -> None:
+        self.jobs_processed += 1
+
     def report_job_scanned(self) -> None:
         self.jobs_scanned += 1
 

metadata-ingestion/src/datahub/ingestion/source/aws/sagemaker_processors/jobs.py

@@ -1,3 +1,4 @@
+import logging
 from collections import defaultdict
 from dataclasses import dataclass, field
 from enum import Enum
@@ -49,6 +50,8 @@
 if TYPE_CHECKING:
     from mypy_boto3_sagemaker import SageMakerClient
 
+logger = logging.getLogger(__name__)
+
 JobInfo = TypeVar(
     "JobInfo",
     AutoMlJobInfo,
@@ -274,15 +277,18 @@ def get_job_details(self, job_name: str, job_type: JobType) -> Dict[str, Any]:
         )
 
     def get_workunits(self) -> Iterable[MetadataWorkUnit]:
+        logger.info("Getting all SageMaker jobs")
         jobs = self.get_all_jobs()
 
         processed_jobs: Dict[str, SageMakerJob] = {}
 
+        logger.info("Processing SageMaker jobs")
         # first pass: process jobs and collect datasets used
+        logger.info("first pass: process jobs and collect datasets used")
         for job in jobs:
             job_type = job_type_to_info[job["type"]]
             job_name = job[job_type.list_name_key]
-
+            logger.debug(f"Processing job {job_name} with type {job_type}")
             job_details = self.get_job_details(job_name, job["type"])
 
             processed_job = getattr(self, job_type.processor)(job_details)
@@ -293,6 +299,9 @@ def get_workunits(self) -> Iterable[MetadataWorkUnit]:
         # second pass:
         #   - move output jobs to inputs
         #   - aggregate i/o datasets
+        logger.info(
+            "second pass: move output jobs to inputs and aggregate i/o datasets"
+        )
         for job_urn in sorted(processed_jobs):
             processed_job = processed_jobs[job_urn]
 
@@ -301,6 +310,7 @@ def get_workunits(self) -> Iterable[MetadataWorkUnit]:
 
             all_datasets.update(processed_job.input_datasets)
             all_datasets.update(processed_job.output_datasets)
+            self.report.report_job_processed()
 
         # yield datasets
         for dataset_urn, dataset in all_datasets.items():
@@ -322,6 +332,7 @@ def get_workunits(self) -> Iterable[MetadataWorkUnit]:
             self.report.report_dataset_scanned()
 
         # third pass: construct and yield MCEs
+        logger.info("third pass: construct and yield MCEs")
         for job_urn in sorted(processed_jobs):
             processed_job = processed_jobs[job_urn]
             job_snapshot = processed_job.job_snapshot

metadata-ingestion/src/datahub/ingestion/source/aws/sagemaker_processors/lineage.py

@@ -1,3 +1,4 @@
+import logging
 from collections import defaultdict
 from dataclasses import dataclass, field
 from typing import TYPE_CHECKING, Any, DefaultDict, Dict, List, Set
@@ -6,6 +7,8 @@
     SagemakerSourceReport,
 )
 
+logger = logging.getLogger(__name__)
+
 if TYPE_CHECKING:
     from mypy_boto3_sagemaker import SageMakerClient
     from mypy_boto3_sagemaker.type_defs import (
@@ -88,7 +91,6 @@ def get_all_contexts(self) -> List["ContextSummaryTypeDef"]:
         paginator = self.sagemaker_client.get_paginator("list_contexts")
         for page in paginator.paginate():
             contexts += page["ContextSummaries"]
-
         return contexts
 
     def get_incoming_edges(self, node_arn: str) -> List["AssociationSummaryTypeDef"]:
@@ -225,27 +227,32 @@ def get_lineage(self) -> LineageInfo:
         """
         Get the lineage of all artifacts in SageMaker.
         """
-
+        logger.info("Getting lineage for SageMaker artifacts...")
+        logger.info("Getting all actions")
         for action in self.get_all_actions():
             self.nodes[action["ActionArn"]] = {**action, "node_type": "action"}
+        logger.info("Getting all artifacts")
         for artifact in self.get_all_artifacts():
             self.nodes[artifact["ArtifactArn"]] = {**artifact, "node_type": "artifact"}
+        logger.info("Getting all contexts")
         for context in self.get_all_contexts():
             self.nodes[context["ContextArn"]] = {**context, "node_type": "context"}
 
+        logger.info("Getting lineage for model deployments and model groups")
         for node_arn, node in self.nodes.items():
+            logger.debug(f"Getting lineage for node {node_arn}")
             # get model-endpoint lineage
             if (
                 node["node_type"] == "action"
                 and node.get("ActionType") == "ModelDeployment"
             ):
                 self.get_model_deployment_lineage(node_arn)
-
+                self.report.model_endpoint_lineage += 1
             # get model-group lineage
             if (
                 node["node_type"] == "context"
                 and node.get("ContextType") == "ModelGroup"
             ):
                 self.get_model_group_lineage(node_arn, node)
-
+                self.report.model_group_lineage += 1
         return self.lineage_info