Data Analyst Associate · Domain 3: Importing Data · 5% of the exam
Task 3.1: Explain the approaches for bringing data into Databricks, covering ingestion from S3, data sharing with external systems via Delta Sharing, API-driven data intake, Auto Loader, Lakeflow Connect managed connectors, and Marketplace; Use the Databricks Workspace UI to upload a data file to the platform
COPY INTO, Auto Loader and read_files over S3 through external locations, managed connectors, OpenSharing and Marketplace, API intake, and the file upload page with its limits.
Study it
Bringing data in: S3, Auto Loader, COPY INTO, managed connectors, OpenSharing and Marketplace
Lesson coming
Uploading a file through the workspace UI
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the Databricks documentation page that proves it.
Question 1 · choose 1
A partner delivered a one-time export of 50 GB of Parquet files to an S3 path that is covered by a Unity Catalog external location on which the analyst holds READ FILES. The files will not change, and the analyst wants them in a managed table, using SQL on a SQL warehouse. Which approach fits?
ACREATE TABLE ... AS SELECT * FROM read_files() on the S3 path
BThe Create or modify a table using file upload page
CA Lakeflow Connect managed SaaS connector pointed at the S3 path
DAn OpenSharing recipient created for the partner's account
Show the answer and why
ACREATE TABLE ... AS SELECT * FROM read_files() on the S3 path
Correct
read_files lets SQL users read files in cloud object storage, and a CREATE TABLE AS SELECT over it loads them once into a managed table.
BThe Create or modify a table using file upload page
Incorrect
File upload takes files from the local machine, at most 10 at a time and under 2 GB in total, which a 50 GB export exceeds.
CA Lakeflow Connect managed SaaS connector pointed at the S3 path
Incorrect
Managed SaaS connectors ingest from SaaS applications, not from files in cloud object storage.
DAn OpenSharing recipient created for the partner's account
Incorrect
A recipient is how a provider shares data with someone; it does not load files from an S3 path into a table.
For files in cloud object storage that a Unity Catalog external location covers, the read_files table-valued function reads them in SQL, and CREATE TABLE AS SELECT writes the result into a managed Delta table. The file upload page is for small local files, managed SaaS connectors read SaaS applications, and OpenSharing shares data between organizations.
An analyst has orders.zip, which holds six CSV files of 180 MB each, and two Excel workbooks of 20 MB each on her laptop. She wants managed tables from all of them through the Create or modify a table using file upload page, with a SQL warehouse selected as compute. How should she upload them?
AUpload orders.zip and both workbooks together in one upload
BExtract the six CSV files and upload all eight files in one single upload
CExtract the CSVs and upload them together, then each workbook alone
DUpload orders.zip alone, then upload both workbooks together
Show the answer and why
AUpload orders.zip and both workbooks together in one upload
Incorrect
Compressed files such as zip are not supported, and only one Excel file can be uploaded at a time.
BExtract the six CSV files and upload all eight files in one single upload
Incorrect
Eight files are within the 10-file limit, but only one Excel file can be uploaded at a time.
CExtract the CSVs and upload them together, then each workbook alone
Correct
The six CSVs are under the 10-file and 2 GB limits once extracted, and each Excel file goes in its own upload; a SQL warehouse can read Excel.
DUpload orders.zip alone, then upload both workbooks together
Incorrect
The page does not extract zip files, and two Excel files cannot be uploaded together.
The file upload page creates or overwrites a managed Delta table from CSV, TSV, JSON, Avro, Parquet, Excel or text files: up to 10 files at a time, under 2 GB in total, one Excel file at a time (Excel needs a SQL warehouse, serverless compute or Databricks Runtime 17.1 or above), and no compressed files such as zip or tar.
A team expects millions of new files over time in an S3 location and wants to ingest them incrementally. Which tool does Databricks recommend at that scale?
AThe file upload page
BA one-time CREATE TABLE AS SELECT
CCOPY INTO run on a schedule
DAuto Loader
Show the answer and why
AThe file upload page
Incorrect
File upload is for small local files, up to 10 at a time.
BA one-time CREATE TABLE AS SELECT
Incorrect
A one-time load does not ingest new files incrementally.
CCOPY INTO run on a schedule
Incorrect
COPY INTO suits files in the order of thousands; Auto Loader is recommended for millions.
DAuto Loader
Correct
For millions of files or more over time, Databricks recommends Auto Loader, which discovers files with fewer operations.
For files in the order of thousands over time, COPY INTO works; for millions or more, Databricks recommends Auto Loader, which needs fewer operations to discover files and can split processing into batches.
An application must run SQL statements on a Databricks SQL warehouse through REST calls to load and read data, without a JDBC driver. Which Databricks API is documented for this?
AThe Jobs API
BThe Genie Agents API
CThe Statement Execution API
DThe Lakebase Postgres wire protocol
Show the answer and why
AThe Jobs API
Incorrect
The Jobs API manages and runs jobs, not individual SQL statements on a warehouse.
BThe Genie Agents API
Incorrect
That API sends natural-language questions to a Genie Agent.
CThe Statement Execution API
Correct
The SQL Statement Execution API runs SQL statements on Databricks SQL warehouses through REST.
DThe Lakebase Postgres wire protocol
Incorrect
Lakebase is a Postgres OLTP database, not SQL warehouse access.
The Databricks SQL Statement Execution API 2.0 runs SQL statements on Databricks SQL warehouses through the REST API, which an application can use for API-driven reads and writes.
An analyst wants to upload a PDF contract from her laptop into Unity Catalog so that colleagues with access can open it. Where should it go?
AInto a Unity Catalog volume, with Upload files to a volume
BInto a new table through Create or modify a table using file upload
CInto a Lakebase database
DInto a metric view
Show the answer and why
AInto a Unity Catalog volume, with Upload files to a volume
Correct
Volumes govern non-tabular files of any format, and the UI uploads files to a volume.
BInto a new table through Create or modify a table using file upload
Incorrect
That page creates tables from CSV, TSV, JSON, Avro, Parquet, Excel or text files, not PDFs.
CInto a Lakebase database
Incorrect
Lakebase is an OLTP database; volumes are for governed files.
DInto a metric view
Incorrect
Metric views define business metrics over tables.
Volumes are Unity Catalog objects that govern non-tabular data of any format. From New > Add or upload data > Upload files to a volume, you can upload files of any format through the UI (up to 5 GB per file), with WRITE VOLUME on the volume.
An analyst drops a CSV file onto the Create or modify a table using file upload page while no compute is selected. What can she do before selecting compute?
ANothing; the page rejects files until compute is running
BStage the file now; preview and setup need active compute
CPreview, configure and create the table; compute is used only later for queries
DCreate the table on a group cluster
Show the answer and why
ANothing; the page rejects files until compute is running
Incorrect
Files can be uploaded to the staging area without connecting to compute.
BStage the file now; preview and setup need active compute
Correct
The page allows staging uploads without compute, but preview and table configuration require an active compute resource.
CPreview, configure and create the table; compute is used only later for queries
Incorrect
Previewing and configuring the table need an active compute resource.
DCreate the table on a group cluster
Incorrect
The upload UI supports SQL warehouses, serverless and dedicated compute, not group clusters.
You can upload data to the staging area without connecting to compute, but you must select an active compute resource (a SQL warehouse, serverless compute or dedicated compute) to preview and configure the table.