Amazon SageMaker

概要

Amazon SageMaker は、フルマネージド型の機械学習サービスです。Amazon SageMaker を使用して、データサイエンティストや開発者は、機械学習モデルを構築およびトレーニングした後に、実稼働準備ができたホスト環境にモデルを直接デプロイすることができます。

このインテグレーションを有効にすると、Datadog にすべての SageMaker メトリクスを表示できます。

セットアップ

インストール

Amazon Web Services インテグレーションをまだセットアップしていない場合は、最初にセットアップします。

メトリクスの収集

  1. AWS インテグレーションページで、Metric Collection タブの下にある SageMaker が有効になっていることを確認します。
  2. Datadog - Amazon SageMaker インテグレーションをインストールします。

ログの収集

ログの有効化

Amazon SageMaker から S3 バケットまたは CloudWatch のいずれかにログを送信するよう構成します。

: S3 バケットにログを送る場合は、Target prefixamazon_sagemaker に設定されているかを確認してください。

ログを Datadog に送信する方法

  1. Datadog ログコレクション AWS Lambda 関数 をまだ設定していない場合は、設定を行ってください。

  2. lambda 関数がインストールされたら、AWS コンソールから、Amazon SageMaker ログを含む S3 バケットまたは CloudWatch のロググループに手動でトリガーを追加します。

収集データ

メトリクス

aws.sagemaker.consumed_read_requests_units
(count)
The average number of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.maximum
(count)
The maximum number of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.minimum
(count)
The minimum number of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.p90
(count)
The 90th percentile of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.p95
(count)
The 95th percentile of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.p99
(count)
The 99th percentile of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.sample_count
(count)
The sample count of consumed read units over the specified time period.
aws.sagemaker.consumed_read_requests_units.sum
(count)
The sum of consumed read units over the specified time period.
aws.sagemaker.consumed_write_requests_units
(count)
The average number of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.maximum
(count)
The maximum number of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.minimum
(count)
The minimum number of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.p90
(count)
The 90th percentile of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.p95
(count)
The 95th percentile of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.p99
(count)
The 99th percentile of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.sample_count
(count)
The sample count of consumed write units over the specified time period.
aws.sagemaker.consumed_write_requests_units.sum
(count)
The sum of consumed write units over the specified time period.
aws.sagemaker.endpoints.cpuutilization
(gauge)
The percentage of CPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.endpoints.disk_utilization
(gauge)
The percentage of disk space used by the containers on an instance uses.
Shown as percent
aws.sagemaker.endpoints.gpu_memory_utilization
(gauge)
The percentage of GPU memory used by the containers on an instance.
Shown as percent
aws.sagemaker.endpoints.gpu_utilization
(gauge)
The percentage of GPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.endpoints.loaded_model_count
(count)
The number of models loaded in the containers of the multi-model endpoint. This metric is emitted per instance.
aws.sagemaker.endpoints.loaded_model_count.maximum
(count)
The maximum number of models loaded in the containers of the multi-model endpoint. This metric is emitted per instance.
aws.sagemaker.endpoints.loaded_model_count.minimum
(count)
The minimum number of models loaded in the containers of the multi-model endpoint. This metric is emitted per instance.
aws.sagemaker.endpoints.loaded_model_count.sample_count
(count)
The sample count of models loaded in the containers of the multi-model endpoint. This metric is emitted per instance.
aws.sagemaker.endpoints.loaded_model_count.sum
(count)
The sum of models loaded in the containers of the multi-model endpoint. This metric is emitted per instance.
aws.sagemaker.endpoints.memory_utilization
(gauge)
The percentage of memory that is used by the containers on an instance.
Shown as percent
aws.sagemaker.invocation_4xx_errors
(count)
The average number of InvokeEndpoint requests where the model returned a 4xx HTTP response code.
Shown as request
aws.sagemaker.invocation_4xx_errors.sum
(count)
The sum of the number of InvokeEndpoint requests where the model returned a 4xx HTTP response code.
Shown as request
aws.sagemaker.invocation_5xx_errors
(count)
The average number of InvokeEndpoint requests where the model returned a 5xx HTTP response code.
Shown as request
aws.sagemaker.invocation_5xx_errors.sum
(count)
The sum of the number of InvokeEndpoint requests where the model returned a 5xx HTTP response code.
Shown as request
aws.sagemaker.invocations
(count)
The number of InvokeEndpoint requests sent to a model endpoint.
Shown as request
aws.sagemaker.invocations.maximum
(count)
The maximum of the number of InvokeEndpoint requests sent to a model endpoint.
Shown as request
aws.sagemaker.invocations.minimum
(count)
The minimum of the number of InvokeEndpoint requests sent to a model endpoint.
Shown as request
aws.sagemaker.invocations.sample_count
(count)
The sample count of the number of InvokeEndpoint requests sent to a model endpoint.
Shown as request
aws.sagemaker.invocations_per_instance
(count)
The number of invocations sent to a model normalized by InstanceCount in each ProductionVariant.
aws.sagemaker.invocations_per_instance.sum
(count)
The sum of invocations sent to a model normalized by InstanceCount in each ProductionVariant.
aws.sagemaker.jobs_failed
(count)
The average number of occurrences a single labeling job failed.
Shown as job
aws.sagemaker.jobs_failed.sample_count
(count)
The sample count of occurrences a single labeling job failed.
Shown as job
aws.sagemaker.jobs_failed.sum
(count)
The sum of occurrences a single labeling job failed.
Shown as job
aws.sagemaker.jobs_stopped
(count)
The average number of occurrences a single labeling job was stopped.
Shown as job
aws.sagemaker.jobs_stopped.sample_count
(count)
The sample count of occurrences a single labeling job was stopped.
Shown as job
aws.sagemaker.jobs_stopped.sum
(count)
The sum of occurrences a single labeling job was stopped.
Shown as job
aws.sagemaker.labelingjobs.dataset_objects_auto_annotated
(count)
The average number of dataset objects auto-annotated in a labeling job.
aws.sagemaker.labelingjobs.dataset_objects_auto_annotated.max
(count)
The maximum number of dataset objects auto-annotated in a labeling job.
aws.sagemaker.labelingjobs.dataset_objects_human_annotated
(count)
The average number of dataset objects annotated by a human in a labeling job.
aws.sagemaker.labelingjobs.dataset_objects_human_annotated.max
(count)
The maximum number of dataset objects annotated by a human in a labeling job.
aws.sagemaker.labelingjobs.dataset_objects_labeling_failed
(count)
The number of dataset objects that failed labeling in a labeling job.
aws.sagemaker.labelingjobs.dataset_objects_labeling_failed.max
(count)
The number of dataset objects that failed labeling in a labeling job.
aws.sagemaker.labelingjobs.jobs_succeeded
(count)
The average number of occurrences a single labeling job succeeded.
Shown as job
aws.sagemaker.labelingjobs.jobs_succeeded.sample_count
(count)
The sample count of occurrences a single labeling job succeeded.
Shown as job
aws.sagemaker.labelingjobs.jobs_succeeded.sum
(count)
The sum of occurrences a single labeling job succeeded.
Shown as job
aws.sagemaker.labelingjobs.total_dataset_objects_labeled
(count)
The average number of dataset objects labeled successfully in a labeling job.
aws.sagemaker.labelingjobs.total_dataset_objects_labeled.maximum
(count)
The maximum number of dataset objects labeled successfully in a labeling job.
aws.sagemaker.model_cache_hit
(count)
The number of InvokeEndpoint requests sent to the multi-model endpoint for which the model was already loaded.
Shown as request
aws.sagemaker.model_cache_hit.maximum
(count)
The maximum number of InvokeEndpoint requests sent to the multi-model endpoint for which the model was already loaded.
Shown as request
aws.sagemaker.model_cache_hit.minimum
(count)
The minimum number of InvokeEndpoint requests sent to the multi-model endpoint for which the model was already loaded.
Shown as request
aws.sagemaker.model_cache_hit.sample_count
(count)
The sample count of InvokeEndpoint requests sent to the multi-model endpoint for which the model was already loaded.
Shown as request
aws.sagemaker.model_cache_hit.sum
(count)
The sum of InvokeEndpoint requests sent to the multi-model endpoint for which the model was already loaded.
Shown as request
aws.sagemaker.model_downloading_time
(count)
The interval of time that it takes to download the model from Amazon Simple Storage Service (Amazon S3).
Shown as microsecond
aws.sagemaker.model_downloading_time.maximum
(count)
The maximum interval of time that it takes to download the model from Amazon Simple Storage Service (Amazon S3).
Shown as microsecond
aws.sagemaker.model_downloading_time.minimum
(count)
The minimum interval of time that it takes to download the model from Amazon Simple Storage Service (Amazon S3).
Shown as microsecond
aws.sagemaker.model_downloading_time.sample_count
(count)
The sample count interval of time that it takes to download the model from Amazon Simple Storage Service (Amazon S3).
Shown as microsecond
aws.sagemaker.model_downloading_time.sum
(count)
The sum interval of time that it takes to download the model from Amazon Simple Storage Service (Amazon S3).
Shown as microsecond
aws.sagemaker.model_latency
(count)
The average interval of time taken by a model to respond as viewed from Amazon SageMaker.
Shown as microsecond
aws.sagemaker.model_latency.maximum
(count)
The maximum interval of time taken by a model to respond as viewed from Amazon SageMaker.
Shown as microsecond
aws.sagemaker.model_latency.minimum
(count)
The minimum interval of time taken by a model to respond as viewed from Amazon SageMaker.
Shown as microsecond
aws.sagemaker.model_latency.sample_count
(count)
The sample count interval of time taken by a model to respond as viewed from Amazon SageMaker.
Shown as microsecond
aws.sagemaker.model_latency.sum
(count)
The sum of the interval of time taken by a model to respond as viewed from Amazon SageMaker.
Shown as microsecond
aws.sagemaker.model_loading_time
(count)
The interval of time that it takes to load the model through the container's LoadModel API call.
Shown as microsecond
aws.sagemaker.model_loading_time.maximum
(count)
The maximum interval of time that it takes to load the model through the container's LoadModel API call.
Shown as microsecond
aws.sagemaker.model_loading_time.minimum
(count)
The minimum interval of time that it takes to load the model through the container's LoadModel API call.
Shown as microsecond
aws.sagemaker.model_loading_time.sample_count
(count)
The sample count interval of time that it takes to load the model through the container's LoadModel API call.
Shown as microsecond
aws.sagemaker.model_loading_time.sum
(count)
The sum interval of time that it takes to load the model through the container's LoadModel API call.
Shown as microsecond
aws.sagemaker.model_loading_wait_time
(count)
The interval of time that an invocation request has waited for the target model to be downloaded, or loaded, or both in order to perform inference.
Shown as microsecond
aws.sagemaker.model_loading_wait_time.maximum
(count)
The maximum interval of time that an invocation request has waited for the target model to be downloaded, or loaded, or both in order to perform inference.
Shown as microsecond
aws.sagemaker.model_loading_wait_time.minimum
(count)
The minimum interval of time that an invocation request has waited for the target model to be downloaded, or loaded, or both in order to perform inference.
Shown as microsecond
aws.sagemaker.model_loading_wait_time.sample_count
(count)
The sample count interval of time that an invocation request has waited for the target model to be downloaded, or loaded, or both in order to perform inference.
Shown as microsecond
aws.sagemaker.model_loading_wait_time.sum
(count)
The sum interval of time that an invocation request has waited for the target model to be downloaded, or loaded, or both in order to perform inference.
Shown as microsecond
aws.sagemaker.model_setup_time
(count)
The average time it takes to launch new compute resources for a serverless endpoint.
Shown as microsecond
aws.sagemaker.model_setup_time.maximum
(count)
The maximum interval of time it takes to launch new compute resources for a serverless endpoint.
Shown as microsecond
aws.sagemaker.model_setup_time.minimum
(count)
The minimum interval of time it takes to launch new compute resources for a serverless endpoint.
Shown as microsecond
aws.sagemaker.model_setup_time.sample_count
(count)
The sample_count of the amount of time it takes to launch new compute resources for a serverless endpoint.
Shown as microsecond
aws.sagemaker.model_setup_time.sum
(count)
The total amount of time takes to launch new compute resources for a serverless endpoint.
Shown as microsecond
aws.sagemaker.model_unloading_time
(count)
The interval of time that it takes to unload the model through the container's UnloadModel API call.
Shown as microsecond
aws.sagemaker.model_unloading_time.maximum
(count)
The maximum interval of time that it takes to unload the model through the container's UnloadModel API call.
Shown as microsecond
aws.sagemaker.model_unloading_time.minimum
(count)
The minimum interval of time that it takes to unload the model through the container's UnloadModel API call.
Shown as microsecond
aws.sagemaker.model_unloading_time.sample_count
(count)
The sample count interval of time that it takes to unload the model through the container's UnloadModel API call.
Shown as microsecond
aws.sagemaker.model_unloading_time.sum
(count)
The sum interval of time that it takes to unload the model through the container's UnloadModel API call.
Shown as microsecond
aws.sagemaker.modelbuildingpipeline.execution_duration
(count)
The average duration in milliseconds that the pipeline execution ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.execution_duration.maximum
(count)
The maximum duration in milliseconds that the pipeline execution ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.execution_duration.minimum
(count)
The minimum duration in milliseconds that the pipeline execution ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.execution_duration.sample_count
(count)
The sample count duration in milliseconds that the pipeline execution ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.execution_duration.sum
(count)
The sum duration in milliseconds that the pipeline execution ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.execution_failed
(count)
The average number of steps that failed.
aws.sagemaker.modelbuildingpipeline.execution_failed.sum
(count)
The sum of steps that failed.
aws.sagemaker.modelbuildingpipeline.execution_started
(count)
The average number of pipeline executions that started.
aws.sagemaker.modelbuildingpipeline.execution_started.sum
(count)
The sum of pipeline executions that started.
aws.sagemaker.modelbuildingpipeline.execution_stopped
(count)
The average number of pipeline executions that stopped.
aws.sagemaker.modelbuildingpipeline.execution_stopped.sum
(count)
The sum of pipeline executions that stopped.
aws.sagemaker.modelbuildingpipeline.execution_succeeded
(count)
The average number of pipeline executions that succeeded.
aws.sagemaker.modelbuildingpipeline.execution_succeeded.sum
(count)
The sum of pipeline executions that succeeded.
aws.sagemaker.modelbuildingpipeline.step_duration
(count)
The average duration in milliseconds that the step ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.step_duration.maximum
(count)
The maximum duration in milliseconds that the step ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.step_duration.minimum
(count)
The minimum duration in milliseconds that the step ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.step_duration.sample_count
(count)
The sample count duration in milliseconds that the step ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.step_duration.sum
(count)
The sum duration in milliseconds that the step ran.
Shown as millisecond
aws.sagemaker.modelbuildingpipeline.step_failed
(count)
The average number of steps that failed.
aws.sagemaker.modelbuildingpipeline.step_failed.sum
(count)
The sum of steps that failed.
aws.sagemaker.modelbuildingpipeline.step_started
(count)
The average number of steps that started.
aws.sagemaker.modelbuildingpipeline.step_started.sum
(count)
The sum of steps that started.
aws.sagemaker.modelbuildingpipeline.step_stopped
(count)
The average number of steps that stopped.
aws.sagemaker.modelbuildingpipeline.step_stopped.sum
(count)
The sum of steps that stopped.
aws.sagemaker.modelbuildingpipeline.step_succeeded
(count)
The average number of steps that succeeded.
aws.sagemaker.modelbuildingpipeline.step_succeeded.sum
(count)
The sum of steps that succeeded.
aws.sagemaker.overhead_latency
(count)
The average interval of time added to the time taken to respond to a client request by Amazon SageMaker overheads.
Shown as microsecond
aws.sagemaker.overhead_latency.maximum
(count)
The maximum interval of time added to the time taken to respond to a client request by Amazon SageMaker overheads.
Shown as microsecond
aws.sagemaker.overhead_latency.minimum
(count)
The minimum interval of time added to the time taken to respond to a client request by Amazon SageMaker overheads.
Shown as microsecond
aws.sagemaker.overhead_latency.sample_count
(count)
The sample count of the interval of time added to the time taken to respond to a client request by Amazon SageMaker overheads.
Shown as microsecond
aws.sagemaker.overhead_latency.sum
(count)
The sum of the interval of time added to the time taken to respond to a client request by Amazon SageMaker overheads.
Shown as microsecond
aws.sagemaker.processingjobs.cpuutilization
(gauge)
The percentage of CPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.processingjobs.disk_utilization
(gauge)
The percentage of disk space used by the containers on an instance uses.
Shown as percent
aws.sagemaker.processingjobs.gpu_memory_utilization
(gauge)
The percentage of GPU memory used by the containers on an instance.
Shown as percent
aws.sagemaker.processingjobs.gpu_utilization
(gauge)
The percentage of GPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.processingjobs.memory_utilization
(gauge)
The percentage of memory that is used by the containers on an instance.
Shown as percent
aws.sagemaker.tasks_returned
(count)
The average number of occurrences a single task was returned.
aws.sagemaker.tasks_returned.sample_count
(count)
The sample count of occurrences a single task was returned.
aws.sagemaker.tasks_returned.sum
(count)
The sum of occurrences a single task was returned.
aws.sagemaker.trainingjobs.cpuutilization
(gauge)
The percentage of CPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.trainingjobs.disk_utilization
(gauge)
The percentage of disk space used by the containers on an instance uses.
Shown as percent
aws.sagemaker.trainingjobs.gpu_memory_utilization
(gauge)
The percentage of GPU memory used by the containers on an instance.
Shown as percent
aws.sagemaker.trainingjobs.gpu_utilization
(gauge)
The percentage of GPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.trainingjobs.memory_utilization
(gauge)
The percentage of memory that is used by the containers on an instance.
Shown as percent
aws.sagemaker.transformjobs.cpuutilization
(gauge)
The percentage of CPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.transformjobs.disk_utilization
(gauge)
The percentage of disk space used by the containers on an instance uses.
Shown as percent
aws.sagemaker.transformjobs.gpu_memory_utilization
(gauge)
The percentage of GPU memory used by the containers on an instance.
Shown as percent
aws.sagemaker.transformjobs.gpu_utilization
(gauge)
The percentage of GPU units that are used by the containers on an instance.
Shown as percent
aws.sagemaker.transformjobs.memory_utilization
(gauge)
The percentage of memory that is used by the containers on an instance.
Shown as percent
aws.sagemaker.workteam.active_workers
(count)
The average number of single active workers on a private work team that submitted, released, or declined a task.
aws.sagemaker.workteam.active_workers.sample_count
(count)
The sample count of single active workers on a private work team that submitted, released, or declined a task.
aws.sagemaker.workteam.active_workers.sum
(count)
The sum of single active workers on a private work team that submitted, released, or declined a task.
aws.sagemaker.workteam.tasks_accepted
(count)
The average number of occurrences a single task was accepted by a worker.
aws.sagemaker.workteam.tasks_accepted.sample_count
(count)
The sample count of occurrences a single task was accepted by a worker.
aws.sagemaker.workteam.tasks_accepted.sum
(count)
The sum of occurrences a single task was accepted by a worker.
aws.sagemaker.workteam.tasks_declined
(count)
The average number of occurrences a single task was declined by a worker.
aws.sagemaker.workteam.tasks_declined.sample_count
(count)
The sample count of occurrences a single task was declined by a worker.
aws.sagemaker.workteam.tasks_declined.sum
(count)
The sum of occurrences a single task was declined by a worker.
aws.sagemaker.workteam.tasks_submitted
(count)
The average number of occurrences a single task was submitted/completed by a private worker.
aws.sagemaker.workteam.tasks_submitted.sample_count
(count)
The average number of occurrences a single task was submitted/completed by a private worker.
aws.sagemaker.workteam.tasks_submitted.sum
(count)
The average number of occurrences a single task was submitted/completed by a private worker.
aws.sagemaker.workteam.time_spent
(count)
The average time spent on a task completed by a private worker.
aws.sagemaker.workteam.time_spent.sample_count
(count)
The average time spent on a task completed by a private worker.
aws.sagemaker.workteam.time_spent.sum
(count)
The average time spent on a task completed by a private worker.

イベント

Amazon SageMaker インテグレーションには、イベントは含まれません。

サービスのチェック

Amazon SageMaker インテグレーションには、サービスのチェック機能は含まれません。

トラブルシューティング

ご不明な点は、Datadog のサポートチームまでお問合せください。