Question 1
Pinecrest Research processes research data and distributed collaboration during an approved maintenance window. The investigation finds: A nightly ETL task runs on an always-on interactive cluster and spends most of its budget while idle. Which administrative or engineering action BEST resolves the stated issue?
Explanation
Correct Answer
Use appropriately sized job compute with a finite task lifecycle instead of reserving interactive compute continuously. This targets the initiating failure: A nightly ETL task runs on an always-on interactive cluster and spends most of its budget while idle. For Pinecrest Research and its research data and distributed collaboration, the acceptance check is: A cost report shows compute starts for scheduled work and terminates after the job without paying for unused interactive hours.
Incorrect Answers
Set an appropriate automatic termination interval for the interactive compute resource. That relates to a different Databricks issue: An interactive exploratory cluster remains running for hours after the last user leaves because it lacks an idle termination rule. At Pinecrest Research, the actual problem is: A nightly ETL task runs on an always-on interactive cluster and spends most of its budget while idle. That alternative by itself does not establish the required outcome: A cost report shows compute starts for scheduled work and terminates after the job without paying for unused interactive hours.
Use a right-sized non-GPU compute option for the SQL-only workload. That relates to a different Databricks issue: A developer leaves an expensive GPU cluster running for ordinary SQL transformations that do not use GPU acceleration. At Pinecrest Research, the actual problem is: A nightly ETL task runs on an always-on interactive cluster and spends most of its budget while idle. That alternative by itself does not establish the required outcome: A cost report shows compute starts for scheduled work and terminates after the job without paying for unused interactive hours.
Separate compute access using appropriate permission boundaries and supported isolation settings. That relates to a different Databricks issue: Sensitive workloads from independent teams share unrestricted interactive compute access and can execute one another's notebooks. At Pinecrest Research, the actual problem is: A nightly ETL task runs on an always-on interactive cluster and spends most of its budget while idle. That alternative by itself does not establish the required outcome: A cost report shows compute starts for scheduled work and terminates after the job without paying for unused interactive hours.