Batch ingestion API overview

The Adobe Experience Platform Batch Ingestion API allows you to ingest data into Platform as batch files. Data being ingested can be profile data from a flat file (such as a Parquet file) or data that conforms to a known schema in the Experience Data Model (XDM) registry.

The Batch Ingestion API reference provides additional information on these API calls.

The following diagram outlines the batch ingestion process:

Getting started

The API endpoints used in this guide is part of the Batch Ingestion API. Before continuing, please review the getting started guide for links to related documentation, a guide to reading the sample API calls in this document, and important information regarding required headers that are needed to successfully make calls to any Experience Platform API.

Data Ingestion prerequisites

Batch ingestion best practices

Batch ingestion constraints

Batch data ingestion has some constraints:

Types

When ingesting data, it is important to understand how Experience Data Model (XDM) schemas work. For more information about how XDM field types map to different formats, please read the Schema Registry developer guide.

There is some flexibility when ingesting data - if a type does not match what is in the target schema, the data will be converted to the expressed target type. If it cannot, it will fail the batch with a TypeCompatibilityException.

For example, neither JSON nor CSV has a date or date-time type. As a result, these values are expressed using ISO 8601 formatted strings (“2018-07-10T15:05:59.000-08:00”) or Unix Time formatted in milliseconds (1531263959000) and are converted at ingestion time to the target XDM type.

The table below shows the conversions supported when ingesting data.

Using the API

The Data Ingestion API allows you to ingest data as batches (a unit of data that consists of one or more files to be ingested as a single unit) into Experience Platform in three basic steps:

Create a batch

Before data can be added to a dataset, it must be linked to a batch, which will later be uploaded into a specified dataset.

POST /batches

Request

curl -X POST "https://platform.adobe.io/data/foundation/import/batches" \
  -H "Content-Type: application/json" \
  -H "x-gw-ims-org-id: {ORG_ID}" \
  -H "x-sandbox-name: {SANDBOX_NAME}" \
  -H "Authorization: Bearer {ACCESS_TOKEN}" \
  -H "x-api-key: {API_KEY}"
  -d '{
          "datasetId": "{DATASET_ID}"
      }'

Reponse

{
    "id": "{BATCH_ID}",
    "imsOrg": "{ORG_ID}",
    "updated": 0,
    "status": "loading",
    "created": 0,
    "relatedObjects": [
        {
            "type": "dataSet",
            "id": "{DATASET_ID}"
        }
    ],
    "version": "1.0.0",
    "tags": {},
    "createdUser": "{USER_ID}",
    "updatedUser": "{USER_ID}"
}

File upload

After successfully creating a new batch for uploading, files can then be uploaded to a specific dataset.

You can upload files using the Small File Upload API. However, if your files are too large and the gateway limit is exceeded (such as extended timeouts, requests for body size exceeded, and other constrictions), you can switch over to the Large File Upload API. This API uploads the file in chunks, and stitches data together using the Large File Upload Complete API call.

Small file upload

Once a batch is created, data can be uploaded to a preexisting dataset. The file being uploaded must match its referenced XDM schema.

PUT /batches/{BATCH_ID}/datasets/{DATASET_ID}/files/{FILE_NAME}

Request

curl -X PUT "https://platform.adobe.io/data/foundation/import/batches/{BATCH_ID}/datasets/{DATASET_ID}/files/{FILE_NAME}.parquet" \
  -H "content-type: application/octet-stream" \
  -H "x-gw-ims-org-id: {ORG_ID}" \
  -H "x-sandbox-name: {SANDBOX_NAME}" \
  -H "Authorization: Bearer {ACCESS_TOKEN}" \
  -H "x-api-key: {API_KEY}" \
  --data-binary "@{FILE_PATH_AND_NAME}.parquet"

Reponse

#Status 200 OK, with empty response body

Large file upload - create file

To upload a large file, the file must be split into smaller chunks and uploaded one at a time.

POST /batches/{BATCH_ID}/datasets/{DATASET_ID}/files/{FILE_NAME}?action=initialize

Request

curl -X POST "https://platform.adobe.io/data/foundation/import/batches/{BATCH_ID}/datasets/{DATASET_ID}/files/part1=a/part2=b/{FILE_NAME}.parquet?action=initialize" \
  -H "x-gw-ims-org-id: {ORG_ID}" \
  -H "x-sandbox-name: {SANDBOX_NAME}" \
  -H "Authorization: Bearer {ACCESS_TOKEN}" \
  -H "x-api-key: {API_KEY}"

Reponse

#Status 201 CREATED, with empty response body

Large file upload - upload subsequent parts

After the file has been created, all subsequent chunks can be uploaded by making repeated PATCH requests, one for each section of the file.

PATCH /batches/{BATCH_ID}/datasets/{DATASET_ID}/files/{FILE_NAME}

Request

curl -X PATCH "https://platform.adobe.io/data/foundation/import/batches/{BATCH_ID}/datasets/{DATASET_ID}/files/part1=a/part2=b/{FILE_NAME}.parquet" \
  -H "content-type: application/octet-stream" \
  -H "x-gw-ims-org-id: {ORG_ID}" \
  -H "x-sandbox-name: {SANDBOX_NAME}" \
  -H "Authorization: Bearer {ACCESS_TOKEN}" \
  -H "x-api-key: {API_KEY}" \
  -H "Content-Range: bytes {CONTENT_RANGE}" \
  --data-binary "@{FILE_PATH_AND_NAME}.parquet"

Reponse

#Status 200 OK, with empty response

Signal batch completion

After all files have been uploaded to the batch, the batch can be signaled for completion. By doing this, the Catalog DataSetFile entries are created for the completed files and associated with the batch generated above. The Catalog batch is then marked as successful, which triggers downstream flows to ingest the available data.

Request

POST /batches/{BATCH_ID}?action=COMPLETE

curl -X POST "https://platform.adobe.io/data/foundation/import/batches/{BATCH_ID}?action=COMPLETE" \
-H "x-gw-ims-org-id: {ORG_ID}" \
-H "x-sandbox-name: {SANDBOX_NAME}" \
-H "Authorization: Bearer {ACCESS_TOKEN}" \
-H "x-api-key: {API_KEY}"

Reponse

#Status 200 OK, with empty response

Check batch status

While waiting for the files to uploaded to the batch, the batch’s status can be checked to see its progress.

API format

GET /batch/{BATCH_ID}

Request

curl GET "https://platform.adobe.io/data/foundation/catalog/batch/{BATCH_ID}" \
  -H "Authorization: Bearer {ACCESS_TOKEN}" \
  -H "x-gw-ims-org-id: {ORG_ID}" \
  -H "x-sandbox-name: {SANDBOX_NAME}" \
  -H "x-api-key: {API_KEY}"

Reponse

{
    "{BATCH_ID}": {
        "imsOrg": "{ORG_ID}",
        "created": 1494349962314,
        "createdClient": "MCDPCatalogService",
        "createdUser": "{USER_ID}",
        "updatedUser": "{USER_ID}",
        "updated": 1494349963467,
        "externalId": "{EXTERNAL_ID}",
        "status": "success",
        "errors": [
            {
                "code": "err-1494349963436"
            }
        ],
        "version": "1.0.3",
        "availableDates": {
            "startDate": 1337,
            "endDate": 4000
        },
        "relatedObjects": [
            {
                "type": "batch",
                "id": "foo_batch"
            },
            {
                "type": "connection",
                "id": "foo_connection"
            },
            {
                "type": "connector",
                "id": "foo_connector"
            },
            {
                "type": "dataSet",
                "id": "foo_dataSet"
            },
            {
                "type": "dataSetView",
                "id": "foo_dataSetView"
            },
            {
                "type": "dataSetFile",
                "id": "foo_dataSetFile"
            },
            {
                "type": "expressionBlock",
                "id": "foo_expressionBlock"
            },
            {
                "type": "service",
                "id": "foo_service"
            },
            {
                "type": "serviceDefinition",
                "id": "foo_serviceDefinition"
            }
        ],
        "metrics": {
            "foo": 1337
        },
        "tags": {
            "foo_bar": [
                "stuff"
            ],
            "bar_foo": [
                "woo",
                "baz"
            ],
            "foo/bar/foo-bar": [
                "weehaw",
                "wee:haw"
            ]
        },
        "inputFormat": {
            "format": "parquet",
            "delimiter": ".",
            "quote": "`",
            "escape": "\\",
            "nullMarker": "",
            "header": "true",
            "charset": "UTF-8"
        }
    }
}

The "status" field is what shows the current status of the batch requested. The batches can have one of the following states:

Batch ingestion statuses