AWS Security ChangesHomeSearch

AWS iot-sitewise: Scenario Discovery pipeline docs: S3 mounts, storage, IAM

Service: iot-sitewise · 2026-11-05 · Documentation medium

File: iot-sitewise/latest/userguide/sd-workflows.md · Type: iam

Summary

Restructures the Scenario Discovery data pipeline documentation, adding new sections on pipeline execution storage (shared storage, task ephemeral storage, and read-only S3 mounts), service-defined environment variables with reserved AWS_ prefix rules, S3 mount configuration parameters and permissions, and task ephemeral storage configuration. It also documents the task execution role trust policy (iotsitewise.amazonaws.com), required s3:ListBucket/s3:GetObject and kms:Decrypt permissions, and that ephemeral volumes are encrypted at rest.

Security assessment

The diff adds security-relevant guidance rather than fixing a specific vulnerability: it documents that task ephemeral volumes are encrypted at rest, that S3 mounts are read-only, that the task execution role must allow s3:ListBucket/s3:GetObject through the access point plus kms:Decrypt for SSE-KMS buckets, and that AWS_-prefixed environment variable names are reserved and rejected to avoid conflicts with service-defined variables. These are hardening/best-practice docs with no referenced CVE or incident, so security_issue_related is false while adds_security_documentation is true.

Evidence

  * **Encryption.** Volumes are encrypted at rest.

Diff

diff --git a/iot-sitewise/latest/userguide/sd-workflows.md b/iot-sitewise/latest/userguide/sd-workflows.md
index e0d462c..a2dab0b 100644
--- a//iot-sitewise/latest/userguide/sd-workflows.md
+++ b//iot-sitewise/latest/userguide/sd-workflows.md
@@ -7 +7 @@
-Important data pipeline concepts in Scenario DiscoveryPrerequisites to using containers, tasks and pipelinesProcess and ingest data using tasks and pipelines
+Important data pipeline concepts in Scenario DiscoveryConfigure and run data pipelines
@@ -13 +13,3 @@ Important data pipeline concepts in Scenario DiscoveryPrerequisites to using con
-This section covers important concepts related to data pipelines within Scenario Discovery. Data pipelines, and their constituent tasks, give you a flexible, containerized method to perform arbitrary tasks on your data and to store, retrieve, or interact with the Scenario Discovery feature. You bind pipelines and tasks to a workspace and define them there. Understanding these concepts enables you to use the system more efficiently.
+This section covers important concepts related to data pipelines within Scenario Discovery. Data pipelines provide a managed orchestration and compute layer for workloads such as data processing, simulation, data ingestion, and enrichment jobs. You focus on writing your business logic as containerized tasks, and Scenario Discovery provisions and manages the underlying infrastructure. You bind pipelines and tasks to a workspace and define them there. Understanding these concepts enables you to use the system more efficiently.
+
+This section describes concepts only. For setup and step-by-step instructions, see Configure and run data pipelines.
@@ -17 +19 @@ This section covers important concepts related to data pipelines within Scenario
-A pipeline is a series of data tasks you assemble in a directed acyclic graph (DAG). Tasks, represented as nodes in the graph, are individual units of logic that perform some kind of processing on their input data and can provide output data or metadata for use by subsequent nodes. A pipeline can contain one or many nodes. You can assemble nodes in parallel, in series, or in combinations of both. Execution proceeds level by level: all tasks at a single level run until they reach a terminal state before the system proceeds to the next layer in the DAG. If any single node fails in a level, execution for the complete DAG stops.
+A pipeline is a series of data tasks you assemble in a directed acyclic graph (DAG). Tasks, represented as nodes in the graph, are individual units of logic. Each task performs some kind of processing on its input data and can provide output data or metadata for use by subsequent nodes. A pipeline can contain one or many nodes. You can assemble nodes in parallel, in series, or in combinations of both. Execution proceeds level by level: all tasks at a single level run until they reach a terminal state before the system proceeds to the next layer in the DAG. If any single node fails in a level, execution for the complete DAG stops.
@@ -21 +23,19 @@ A pipeline is a series of data tasks you assemble in a directed acyclic graph (D
-Enrichment is the process by which Scenario Discovery analyzes your onboarded video data using domain-appropriate AI models to automatically extract meaningful behaviors and objects. During enrichment, Scenario Discovery recognizes common industry video data formats and applies AI-powered understanding to identify objects, events, environmental conditions, and relationships within your data. Once enrichment is complete, your data becomes fully searchable and discoverable, enabling you to use natural-language queries, receive proactive recommendations, and leverage the full suite of Scenario Discovery curation capabilities.
+Enrichment is the process by which Scenario Discovery analyzes your onboarded video data using domain-appropriate AI models to automatically extract meaningful behaviors and objects. During enrichment, Scenario Discovery recognizes common industry video data formats and applies AI-powered understanding to identify objects, events, environmental conditions, and relationships within your data. After enrichment is complete, your data becomes fully searchable and discoverable. You can then run natural-language queries, receive proactive recommendations, and use the full suite of Scenario Discovery curation capabilities.
+
+### Pipeline execution storage
+
+Pipeline tasks can use two types of storage: shared storage and task ephemeral storage. Tasks can also read data directly from Amazon S3 through S3 mounts. Each option is exposed to the task container as a file system path, so your application can access data by using standard file system operations.
+
+#### Shared storage
+
+Pipeline tasks have access to shared ephemeral storage, mounted at `/var/tmp/data`. The path is also available in the `AWS_PIPELINE_SHARED_STORAGE_ROOT` environment variable. This storage is available during pipeline execution and provides shared access between tasks in the same pipeline. Data in shared storage is temporary and does not persist after the pipeline execution reaches a terminal state (succeeded, failed, or canceled).
+
+#### Task ephemeral storage
+
+Task ephemeral storage is dedicated, high-throughput local storage for intermediate processing within a single compute node. It is attached to the compute node for the duration of that compute node run. You specify the volume size and a storage class, which determines the IOPS and throughput of the volume. Scenario Discovery mounts the volume at `/var/tmp/scratch` and sets the `AWS_TASK_LOCAL_STORAGE_ROOT` environment variable. The volume is encrypted, is provisioned empty for each compute node run, and is deleted when the compute node completes. For configuration steps, see Configure task ephemeral storage.
+
+#### S3 mounts
+
+An S3 mount makes the objects in an Amazon S3 access point accessible as read-only files in the task container, at `$AWS_PIPELINE_SHARED_STORAGE_ROOT/<relativePath>`. An S3 mount isn't a storage layer: data remains in Amazon S3 and is retrieved as your application reads it. You can optionally restrict the mount to a key prefix. For configuration steps, see Mount Amazon S3 data into tasks.
+
+#### Comparison
@@ -23 +43 @@ Enrichment is the process by which Scenario Discovery analyzes your onboarded vi
-### Pipeline shared ephemeral storage
+The following table compares the three ways a task can access data.
@@ -25 +45,8 @@ Enrichment is the process by which Scenario Discovery analyzes your onboarded vi
-Pipeline tasks have access to shared ephemeral storage, mounted at `/var/tmp/data`. This storage is available during pipeline execution and provides shared access between tasks in the same pipeline. Data in shared storage is temporary and does not persist after the pipeline completes.
+Characteristic | Shared storage | Task ephemeral storage | S3 mount  
+---|---|---|---  
+Path in container | `$AWS_PIPELINE_SHARED_STORAGE_ROOT` (`/var/tmp/data`) | `$AWS_TASK_LOCAL_STORAGE_ROOT` (`/var/tmp/scratch`) | `$AWS_PIPELINE_SHARED_STORAGE_ROOT/<relativePath>`  
+Scope | Shared by all compute nodes in a pipeline execution | Dedicated to a single compute node | The compute nodes for which the mount is configured  
+Access | Read/write | Read/write | Read-only  
+Configuration | Provided automatically | Optional; configured on the task | Optional; configured on the task, or on individual compute nodes when you start a pipeline execution  
+Lifecycle | Deleted when the pipeline execution reaches a terminal state | Deleted when the compute node completes | Unmounted when the compute node completes. The source objects in Amazon S3 are unchanged.  
+Recommended use | Exchanging data between compute nodes in a pipeline execution | High-performance temporary storage for a single compute node | Read-only access to source data in Amazon S3 without copying it  
@@ -29 +56 @@ Pipeline tasks have access to shared ephemeral storage, mounted at `/var/tmp/dat
-Tasks are the base functional unit of a Scenario Discovery data pipeline. A task is containerized software logic you use to process, retrieve, or otherwise interact with Scenario Discovery functions. You package tasks as container images, and Scenario Discovery requires the images to be accessible from AWS Elastic Container Registry ([https://aws.amazon.com/ecr/](https://aws.amazon.com/ecr/)).
+Tasks are the base functional unit of a Scenario Discovery data pipeline. A task is containerized software logic you use to process, retrieve, or otherwise interact with Scenario Discovery functions. You package tasks as container images, and Scenario Discovery requires the images to be accessible from Amazon Elastic Container Registry (Amazon ECR) ([https://aws.amazon.com/ecr/](https://aws.amazon.com/ecr/)).
@@ -34,0 +62,18 @@ You use environment variables in tasks and pipelines to pass parameters to conta
+Environment variable names must start with a letter or underscore and contain only letters, digits, and underscores (maximum 255 characters). Values can be up to 1,024 characters. Each environment variable map can contain up to 20 entries. Names that begin with `AWS_` are reserved for service-defined environment variables.
+
+#### Service-defined environment variables
+
+Scenario Discovery sets the following environment variables in every task container. We recommend that you reference storage locations through these variables rather than through hard-coded paths.
+
+Variable | Value | Availability  
+---|---|---  
+`AWS_PIPELINE_SHARED_STORAGE_ROOT` | Path to shared storage (`/var/tmp/data`) | Always  
+`AWS_TASK_LOCAL_STORAGE_ROOT` | Path to task ephemeral storage (`/var/tmp/scratch`) | Only when the task configuration includes `ephemeralStorageConfiguration`  
+`AWS_PIPELINE_EXECUTION_ID` | ID of the current pipeline execution | Always  
+`AWS_PIPELINE_NAME` | Name of the pipeline | Always  
+`AWS_WORKSPACE_NAME` | Name of the workspace | Always  
+`AWS_TASK_NAME` | Name of the task that the compute node runs | Always  
+`AWS_COMPUTE_NODE_NAME` | Name of the compute node in the pipeline definition | Always  
+  
+Scenario Discovery rejects any task, pipeline, or execution environment variable whose name begins with `AWS_` (case-insensitive). This ensures that the environment variables you define don't conflict with service-defined environment variables.
+
@@ -37 +82 @@ You use environment variables in tasks and pipelines to pass parameters to conta
-The system resolves environment variables at runtime and records them in the execution instance. Precedence follows this pattern: pipeline environment variables take precedence over task environment variables, and execution-time changes take precedence over creation-time definitions. For example, suppose you define a variable called `S3_PATH` in a task called "load_data." Next you define a pipeline called "ingest_processing" that contains this task and defines the same `S3_PATH` environment variable. The final value used at runtime is the value from the pipeline's `S3_PATH`, not the task's `S3_PATH` value. If at execution time you change the task's `S3_PATH`, this takes precedence over the original pipeline's `S3_PATH`. And if you modify `S3_PATH` at the pipeline level at execution time, this takes precedence over any other definition of this environment variable.
+The system resolves environment variables at runtime and records them in the execution instance. Precedence follows this pattern: pipeline environment variables take precedence over task environment variables, and execution-time changes take precedence over creation-time definitions. For example, suppose you define a variable called `S3_PATH` in a task called "load_data." Next you define a pipeline called "ingest_processing" that contains this task and defines the same `S3_PATH` environment variable. The final value used at runtime is the value from the pipeline's `S3_PATH`, not the task's `S3_PATH` value. If at execution time you change the task's `S3_PATH`, this takes precedence over the original pipeline's `S3_PATH`. And if you modify `S3_PATH` at the pipeline level at execution time, this takes precedence over any other definition of this environment variable. Service-defined environment variables are always set by Scenario Discovery and can't be overridden.
@@ -39 +84 @@ The system resolves environment variables at runtime and records them in the exe
-## Prerequisites to using containers, tasks and pipelines
+## Configure and run data pipelines
@@ -41 +86 @@ The system resolves environment variables at runtime and records them in the exe
-### Transfer data from Amazon S3 to pipeline storage
+This section describes how to set up and run data pipelines. It covers the prerequisites, the configuration of S3 mounts and task ephemeral storage, and the steps to create and run tasks and pipelines in the AWS IoT SiteWise console.
@@ -43 +88 @@ The system resolves environment variables at runtime and records them in the exe
-Pipeline tasks access data from shared storage mounted at `/var/tmp/data`. You must transfer your source data from Amazon S3 to this location before your pipeline tasks can process it.
+### Prerequisites to using containers, tasks and pipelines
@@ -45 +90 @@ Pipeline tasks access data from shared storage mounted at `/var/tmp/data`. You m
-### Ingest your data from Amazon S3 into AWS IoT SiteWise
+Complete the following prerequisites before you process and ingest data with tasks and pipelines.
@@ -47 +92 @@ Pipeline tasks access data from shared storage mounted at `/var/tmp/data`. You m
-You use the `CreateBulkImportJob` API to ingest MP4 video, Parquet telemetry, and OpenLABEL annotation files from Amazon S3 into AWS IoT SiteWise for Scenario Discovery. Before you can call this API, you must enable the warm tier on your account (a one-time setup per account/region) and configure two IAM roles: one that AWS IoT SiteWise assumes to read your S3 objects and write error reports, and one that your calling identity uses to invoke the API and pass the service role. You define consistent aliases for each file type so Scenario Discovery can correlate video, telemetry, and annotations from the same event. You can invoke the API through the AWS CLI, the boto3 Python SDK, or raw HTTPS (signed with SigV4), and then verify job completion by polling `DescribeBulkImportJob` for a terminal status. For step-by-step instructions, IAM policy examples, request payloads, and code samples, see [CreateBulkImportJob for Scenario Discovery](./sd-bulk-import.html).
+#### Ingest your data from Amazon S3 into AWS IoT SiteWise
@@ -49 +94 @@ You use the `CreateBulkImportJob` API to ingest MP4 video, Parquet telemetry, an
-### Enrich videos for search
+You use the `CreateBulkImportJob` API to ingest MP4 video, Parquet telemetry, and OpenLABEL annotation files from Amazon S3 into AWS IoT SiteWise for Scenario Discovery. Before you can call this API, you must enable the warm tier on your account (a one-time setup per account and Region) and configure two IAM roles. AWS IoT SiteWise assumes the first role to read your S3 objects and write error reports. Your calling identity uses the second role to invoke the API and pass the service role. You define consistent aliases for each file type so Scenario Discovery can correlate video, telemetry, and annotations from the same event. You can invoke the API through the AWS CLI, the boto3 Python SDK, or raw HTTPS (signed with SigV4), and then verify job completion by polling `DescribeBulkImportJob` for a terminal status. For step-by-step instructions, IAM policy examples, request payloads, and code samples, see [CreateBulkImportJob for Scenario Discovery](./sd-bulk-import.html).
@@ -51 +96 @@ You use the `CreateBulkImportJob` API to ingest MP4 video, Parquet telemetry, an
-You use the `CreateEnrichmentJob` API to analyze video time-series data already ingested into AWS IoT SiteWise and generate embeddings that enable semantic (natural-language) search over events in that video. This is the natural follow-on to `CreateBulkImportJob`: after you land MP4 video in a Scenario Discovery dataset, you call `CreateEnrichmentJob` to make that video queryable. Before you submit the job, you need an active Scenario Discovery workspace and dataset containing the target video, and your calling identity must have AWS IoT SiteWise enrichment-job permissions (plus KMS decrypt access if the workspace uses a customer managed key). You scope each job to a single video time series — identified by its property alias — and define a time window to process. You can invoke the API through the AWS CLI, the boto3 Python SDK, or raw HTTPS (signed with SigV4), and then verify completion by polling `DescribeEnrichmentJob` until it reaches a terminal status. For step-by-step instructions, IAM policy examples, request payloads, and code samples, see [CreateEnrichmentJob for Scenario Discovery](./sd-enrichment.html).
+#### Enrich videos for search
@@ -53 +98 @@ You use the `CreateEnrichmentJob` API to analyze video time-series data already
-## Process and ingest data using tasks and pipelines
+You use the `CreateEnrichmentJob` API to analyze video time-series data already ingested into AWS IoT SiteWise and generate embeddings that enable semantic (natural-language) search over events in that video. This is the natural follow-on to `CreateBulkImportJob`: after you land MP4 video in a Scenario Discovery dataset, you call `CreateEnrichmentJob` to make that video queryable. Before you submit the job, you need an active Scenario Discovery workspace and dataset containing the target video, and your calling identity must have AWS IoT SiteWise enrichment-job permissions (plus AWS KMS decrypt access if the workspace uses a customer managed key). You scope each job to a single video time series — identified by its property alias — and define a time window to process. You can invoke the API through the AWS CLI, the boto3 Python SDK, or raw HTTPS (signed with SigV4), and then verify completion by polling `DescribeEnrichmentJob` until it reaches a terminal status. For step-by-step instructions, IAM policy examples, request payloads, and code samples, see [CreateEnrichmentJob for Scenario Discovery](./sd-enrichment.html).
@@ -55 +100 @@ You use the `CreateEnrichmentJob` API to analyze video time-series data already
-Scenario Discovery uses tasks and pipelines to process and ingest your data. A task runs a container (your custom code) on managed compute. A pipeline chains multiple tasks together into an automated workflow. This section walks you through creating tasks, building a pipeline, running it, and verifying your data was ingested using examples.
+### Mount Amazon S3 data into tasks
@@ -57 +102 @@ Scenario Discovery uses tasks and pipelines to process and ingest your data. A t
-### Step 1: Navigate to the Tasks tab
+An S3 mount makes the contents of an Amazon S3 access point accessible in the task container. The mount path is the shared storage root followed by the `relativePath` that you specify:
@@ -59 +103,0 @@ Scenario Discovery uses tasks and pipelines to process and ingest your data. A t
-From your workspace, choose the **Tasks** tab. This is where you manage all your container tasks. If you haven't created any tasks yet, the list is empty.
@@ -61 +105 @@ From your workspace, choose the **Tasks** tab. This is where you manage all your
-![Tasks tab in your workspace](/images/iot-sitewise/latest/userguide/images/sd-image11.png)
+    <mount path> = $AWS_PIPELINE_SHARED_STORAGE_ROOT/<relativePath>
@@ -63 +107 @@ From your workspace, choose the **Tasks** tab. This is where you manage all your
-### Step 2: Create a new task
+For example, with `relativePath` set to `inputs/raw-data`, the data appears at `$AWS_PIPELINE_SHARED_STORAGE_ROOT/inputs/raw-data` (`/var/tmp/data/inputs/raw-data`). Your application reads the objects by using standard file operations, without calling Amazon S3 APIs. Scenario Discovery establishes all mounts before the container starts.
@@ -65 +109 @@ From your workspace, choose the **Tasks** tab. This is where you manage all your
-Choose **Create task** and fill in the basic details: give your task a name (for example, "S3 to EFS"), add an optional description, and select "Container" as the task type.
+###### Note
@@ -67 +111 @@ Choose **Create task** and fill in the basic details: give your task a name (for
-![Creating a new task with name and type](/images/iot-sitewise/latest/userguide/images/sd-image12.png)
+S3 mounts are currently available only through the Scenario Discovery API. You can't configure S3 mounts in the AWS IoT SiteWise console.
@@ -69 +113 @@ Choose **Create task** and fill in the basic details: give your task a name (for
-### Step 3: Configure the container image and execution role
+#### Where you configure mounts
@@ -71 +115,4 @@ Choose **Create task** and fill in the basic details: give your task a name (for
-Point the task to your container image by entering its ECR URI (your image address in Amazon Elastic Container Registry). Then select the IAM execution role that grants the container permission to access your AWS resources like S3 buckets.
+Level | Operation | Field | Accessible to  
+---|---|---|---  
+Task | `CreateTask`, `UpdateTask` | `taskConfiguration.containerTaskConfiguration.mounts` | Compute nodes that run the task  
+Compute node, per execution | `StartPipelineExecution` | `executionMountOverrides.computeNodes.<computeNodeName>` | The specified compute node, for that execution only  
@@ -73 +120 @@ Point the task to your container image by entering its ECR URI (your image addre
-![Configuring the container image ECR URI](/images/iot-sitewise/latest/userguide/images/sd-image13.png)
+You can't configure mounts in the pipeline definition, and you can't define a single execution-level mount that applies to all compute nodes. Execution-time mounts are specified for individual compute nodes.
@@ -75 +122 @@ Point the task to your container image by entering its ECR URI (your image addre
-![Selecting the IAM execution role](/images/iot-sitewise/latest/userguide/images/sd-image14.png)
+Scenario Discovery matches execution overrides to task-defined mounts by name:
@@ -77 +124 @@ Point the task to your container image by entering its ECR URI (your image addre
-### Step 4: Set processing parameters
+  * An execution override with the same name as a task-defined mount overrides that mount for that compute node and execution.
@@ -79 +126 @@ Point the task to your container image by entering its ECR URI (your image addre
-Choose how much compute your task needs. Select "Generic compute" for standard CPU workloads or "Hardware accelerated" for GPU tasks. Pick a processing unit size (for example, 16 units), set the startup command (for example, "python main.py"), and specify a timeout in seconds (for example, 3600 for one hour).
+  * An execution override with a different name is added as a separate mount.
@@ -81 +127,0 @@ Choose how much compute your task needs. Select "Generic compute" for standard C
-![Setting compute type and processing parameters](/images/iot-sitewise/latest/userguide/images/sd-image15.png)
@@ -83 +128,0 @@ Choose how much compute your task needs. Select "Generic compute" for standard C
-![Configuring startup command and timeout](/images/iot-sitewise/latest/userguide/images/sd-image16.png)
@@ -85 +129,0 @@ Choose how much compute your task needs. Select "Generic compute" for standard C
-### Step 5: Add environment variables
@@ -87 +131 @@ Choose how much compute your task needs. Select "Generic compute" for standard C
-Add any environment variables your container needs at runtime. Common examples include the AWS region, S3 bucket paths, workspace name, and role ARNs. These are key-value pairs that configure how your container behaves.
+For example, suppose a task defines a mount named `rawdata`. When you start an execution, you specify two mounts for the compute node that runs this task: `rawdata`, which points to a different prefix, and `sensordata`. For that execution, the compute node uses your `rawdata` mount instead of the task-defined mount, and also has the `sensordata` mount.
@@ -89 +133 @@ Add any environment variables your container needs at runtime. Common examples i
-![Adding environment variables to your task](/images/iot-sitewise/latest/userguide/images/sd-image17.png)
+Compute nodes that aren't specified in `executionMountOverrides` use their task-defined mounts without changes. Each key under `computeNodes` must match a `computeNodeName` in the pipeline definition, not a task name.
@@ -91 +135 @@ Add any environment variables your container needs at runtime. Common examples i
-### Step 6: Review the created task
+#### Mount parameters
@@ -93 +137,7 @@ Add any environment variables your container needs at runtime. Common examples i
-Choose **Create task** to finish. A green banner confirms that the task was created successfully. You can review all your settings on the task details page and use the Edit or Delete buttons if you need to make changes.
+Parameter | Required | Description | Constraints  
+---|---|---|---  
+`name` | Yes | A name that identifies the mount. Scenario Discovery also uses this value to match an execution override to a task-defined mount. | Maximum 64 characters. Must be unique within the list.  
+`relativePath` | Yes | The path at which the data is mounted, relative to the shared storage root (`$AWS_PIPELINE_SHARED_STORAGE_ROOT`). | Maximum 1,024 characters. Must be a relative path: it can't begin or end with `/`, contain `//`, or contain `.` or `..` segments. Must be unique within the list.  
+`source.s3AccessPoint.accessPointArn` | Yes | The Amazon Resource Name (ARN) of the Amazon S3 access point to mount. | Must be an S3 access point ARN. Bucket names and bucket ARNs aren't accepted.  
+`source.s3AccessPoint.prefix` | No | An object key prefix. If specified, only objects whose keys begin with this prefix are accessible. To mount all objects in the access point, omit this parameter. | 1–1,024 characters  
+`storageType` | Yes | The storage in which the mount is created. Currently, mounts can be created only in shared storage. | Must be `SHARED_STORAGE`.  
@@ -95 +145 @@ Choose **Create task** to finish. A green banner confirms that the task was crea
-![Reviewing the created task details](/images/iot-sitewise/latest/userguide/images/sd-image18.png)
+Each compute node supports a maximum of 5 mounts, after task-defined mounts and execution overrides are combined.
@@ -97 +147 @@ Choose **Create task** to finish. A green banner confirms that the task was crea
-### Step 7: Create additional tasks
+#### Behavior
@@ -99 +149 @@ Choose **Create task** to finish. A green banner confirms that the task was crea
-Create additional tasks as needed for your pipeline. For example, you might create a separate "ingest" task that handles data ingestion into Scenario Discovery. Each task can use a different container image and configuration.
+  * **Read-only access.** The container can't create, modify, or delete files in a mount. Write output to shared storage, task ephemeral storage, or Amazon S3.
@@ -101 +151 @@ Create additional tasks as needed for your pipeline. For example, you might crea
-![Creating an additional ingestion task](/images/iot-sitewise/latest/userguide/images/sd-image19.png)
+  * **On-demand retrieval.** Objects are retrieved from Amazon S3 when your application reads them. Data isn't copied in advance.
@@ -103 +153 @@ Create additional tasks as needed for your pipeline. For example, you might crea
-### Step 8: Navigate to the Pipelines tab
+  * **Scope.** A mount is accessible only to the compute nodes for which it is configured. To make the same data available to multiple compute nodes, configure the mount in each corresponding task, or specify it for each compute node in `executionMountOverrides`.
@@ -105 +155 @@ Create additional tasks as needed for your pipeline. For example, you might crea
-Go back to your workspace and choose the **Pipelines** tab. Choose **Create pipeline** to start building a multi-step workflow. Give your pipeline a name (for example, "ingestdatapipeline") and an optional description.
+  * **Pricing.** Scenario Discovery doesn't charge for S3 mounts. Standard Amazon S3 charges for requests and data transfer apply when your tasks read mounted objects.
@@ -107 +156,0 @@ Go back to your workspace and choose the **Pipelines** tab. Choose **Create pipe
-![Creating a new pipeline](/images/iot-sitewise/latest/userguide/images/sd-image20.png)
@@ -109 +157,0 @@ Go back to your workspace and choose the **Pipelines** tab. Choose **Create pipe
-### Step 9: Add compute nodes to the pipeline
@@ -111 +158,0 @@ Go back to your workspace and choose the **Pipelines** tab. Choose **Create pipe
-Choose **\+ Add node** to add processing steps to your pipeline. For each node, give it a name and select which task it should run. You can add multiple nodes to chain tasks together (for example, first transfer data from S3, then ingest it into Scenario Discovery).
@@ -113 +160 @@ Choose **\+ Add node** to add processing steps to your pipeline. For each node,
-![Adding compute nodes to the pipeline](/images/iot-sitewise/latest/userguide/images/sd-image21.png)
+#### Permissions
@@ -115 +162 @@ Choose **\+ Add node** to add processing steps to your pipeline. For each node,
-![Selecting tasks for each node](/images/iot-sitewise/latest/userguide/images/sd-image22.png)
+Scenario Discovery assumes the task's `taskExecutionRole` to read the mounted data. For the trust policy that this role requires, see Step 3: Configure the container image and execution role. In addition, the role must meet these requirements:
@@ -117 +164 @@ Choose **\+ Add node** to add processing steps to your pipeline. For each node,
-### Step 10: Define the pipeline graph and dependencies
+  1. Its permissions policy allows `s3:ListBucket` and `s3:GetObject` through the access point.
@@ -119 +166 @@ Choose **\+ Add node** to add processing steps to your pipeline. For each node,
-The visual editor shows your pipeline as a flowchart. Connect nodes with arrows to define the execution order. For example, the data transfer node runs first, and the ingestion node starts only after it completes. Choose **Create pipeline** when your graph looks correct.
+  2. The access point policy and bucket policy, if any, also allow the role.
@@ -121 +168 @@ The visual editor shows your pipeline as a flowchart. Connect nodes with arrows