Query a Federated Database Instance

Querying Data on S3

When deploying your federated database instance, if you specified an S3 bucket with both read and write permissions or AWS S3 s3:PutObject permission, you can also save your query results in your S3 bucket using $out to S3.

If you successfully create or update an object on your S3 data store, Data Federation returns the latest version of that object for any subsequent read requests and all list operations of the objects also reflect the changes. If your query contains multiple stages, each stage receives the most recent data available from the data store as that stage is processed.

By default, Atlas Data Federation does not return documents in any specific order for queries on Data Federations for S3 data stores. Atlas Data Federation reads the partitions concurrently and the underlying storage response order determines which documents Atlas Data Federation returns first, unless you define order using $sort in your query. For example, if you run the same findOne() query twice, you could see different documents, and if you use $skip, different documents might be skipped if $sort is not used in the query.

You incur "Data Processed" costs for the amount of data that Atlas Data Federation processes to return results for your queries in addition to the "Data Returned" cost for the amount of data that Atlas Data Federation returns. For example, for a 10 GB file, you incur the following "Data Processed" cost in addition to the "Data Returned" cost:

• If you have no partitions, Atlas Data Federation reads the entire file to return results for the query. Therefore, you incur 10 GB of "Data Processed" cost.

• If you have 10 partitions of 1 GB each, Atlas Data Federation targets and reads a single partition. Therefore, you incur 1 GB of "Data Processed" cost.

Querying Data in Your Atlas Cluster

If you query a collection in Atlas Data Federation that is mapped to only one Atlas collection, Atlas Data Federation acts as a proxy and forwards your query to Atlas. When acting as a proxy, Atlas Data Federation doesn't scan data into its virtual collection to proces the query thus improving performance and reducing cost. This optimization is not available for queries on Atlas Data Federation collections that are mapped to multiple Atlas collections.

{
  "stores" : [
    {
      "name" : "atlas-store",
      "provider": "atlas",
      "clusterName": "myCluster",
      "projectId": "5e2211c17a3e5a48f5497de3"
    }
  ],
  "databases" : [
    {
      "name" : "atlas-db",
      "collections" : [
        {
          "name" : "foo",
          "dataSources" : [
            {
              "storeName" : "atlas-store",
              "database" : "myFooData",
              "collection" : "foo"
            }
          ]
        },
        {
          "name" : "barbaz",
          "dataSources" : [
            {
              "storeName" : "atlas-store",
              "database" : "myBarData",
              "collection" : "bar"
            },
            {
              "storeName" : "atlas-store",
              "database" : "myBazData",
              "collection" : "baz"
            }
          ]
        }
      ]
    }
  ]
}

You can also save your query results in your Atlas cluster using $out to Atlas.

If you successfully create or update a document in your collection on the Atlas cluster, Data Federation returns the latest version of that document for any subsequent read requests and all list operations of the collection also reflect the changes. If your query contains multiple stages, each stage receives the most recent data available from the data store as that stage is processed.

Atlas logs your queries against your cluster data in the Atlas cluster audit logs. The log entry for a database user is in the following format:

<SERVICE_NAME>-<CUSTOMER_DATA_LAKE_NAME>-<DATABASE_USER_NAME>

For example, for a database user configured in Atlas as "user" : "CN=atlasDataLake-DataLake0-test_datalake0", log entry in the Atlas cluster audit log resembles the following:

{
  "atype" : "authenticate",
  "ts" : { "$date" : "2022-04-29T13:17:54.020+00:00" },
  "local" : { "ip" : "XXXX", "port" : 27017 },
  "remote" : { "ip" : "XXXXX", "port" : 10844 },
  "users" : [ { "user" : "CN=atlasDataLake-DataLake0-test_datalake0", "db" : "$external" } ],
  "roles" : [ { "role" : "backup", "db" : "admin" }, { "role" : "readWriteAnyDatabase", "db" : "admin" }, { "role" : "clusterMonitor", "db" : "admin" }, { "role" : "enableSharding", "db" : "admin" }, { "role" : "atlasAdmin", "db" : "admin" }, { "role" : "dbAdminAnyDatabase", "db" : "admin" } ],
  "param" : { "user" : "CN=atlasDataLake-DataLake0-test_datalake0", "db" : "$external", "mechanism" : "MONGODB-X509" },
  "result" : 0
}

Querying Data in Your Data Lake Datasets

For queries, Atlas Data Federation uses the partitions that you created on fields during the creation of the Atlas Data Lake pipeline. The order of fields in the partitions is important in the same way as it is for Compound Indexes. Data is optimized for queries by the first field, followed by the second field, and so on. Atlas Data Federation parses the partitions in order; if a query omits a particular partition, Atlas Data Federation is less efficient in making use of any partitions that follow that.

Atlas Data Federation is less performant in supporting queries on fields that don't have partitions. Atlas Data Federation doesn't support queries on fields that were explicitly excluded when creating the Atlas Data Lake pipeline.

Querying Data at a HTTP or HTTPS URL

Data Federation also creates one partition for each URL in your collection. When you connect to your federated database instance and run queries, Data Federation processes your queries against the data and returns the query results.

Running Federated Queries

You can use Atlas Data Federation to query and analyze a unified view of data in your Atlas cluster, S3 bucket, and at your HTTP URL. For federated queries, your federated database instance storage configuration must contain the settings that define:

• Your S3, Atlas, and HTTP data stores.

• Federated Database Instances with virtual collections that map to your S3, Atlas, and HTTP data stores.

You can create or update your federated database instance storage configuration using the Visual Editor or the JSON Editor. For more information, see Create From the UI.

When you connect to your federated database instance and run federated queries, Data Federation combines data from your Atlas cluster, S3 bucket, and HTTP store in virtual databases and collections and returns a union of data in the results.

Troubleshooting

Error: We are currently experiencing increased query processing wait times for Atlas Data Federation. Our Engineering team is investigating. Normal service will resume shortly, please try again.

Atlas Data Federation returns this error only when Atlas Data Federation can't execute queries because of resource contention. We recommend that you run your queries again.