Monitor, Troubleshoot, and Optimize Workloads in Azure Databricks DP-750 Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A company wants to implement log streaming from Azure Databricks to Azure Monitor Log Analytics. They have configured diagnostic settings to stream logs, but they are concerned about the cost of storing logs. They want to retain logs for 30 days for troubleshooting, but they also want to reduce storage costs. What should they do?

   • AUse a separate Log Analytics workspace for each cluster.
   • BSet the retention period of the Log Analytics workspace to 30 days.
   • CDisable diagnostic settings for clusters that are not critical.
   • DStream logs to Azure Storage instead of Log Analytics.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A Spark job that performs a join between two large tables is running slowly. In the Spark UI, the DAG shows a stage with a 'Shuffle Read' that is much larger than the input data. The join key has a highly skewed distribution. Which technique is most effective in reducing the shuffle size and improving performance?

   • AAdd a salt key to the join keys to distribute the skewed keys across more partitions.
   • BIncrease the number of shuffle partitions to 2000.
   • CUse a broadcast join by increasing the autoBroadcastJoinThreshold.
   • DEnable spark.sql.adaptive.skewJoin.enabled to automatically handle skew.
   Reveal Solution Discussion Select an answer first
4. 3expert · hard

   A data engineer is running a Spark notebook that processes a large dataset. The notebook runs slowly, and the Spark UI shows that many tasks are failing with 'java.lang.OutOfMemoryError: Java heap space'. The cluster has 8 workers with 8 GB memory each. The engineer has already increased the executor memory, but the issue persists. What should they do next?

   • AIncrease the number of partitions to reduce the data size per task.
   • BCache the DataFrame to avoid recomputation.
   • CUse a broadcast join to avoid shuffling large data.
   • DEnable speculative execution to re-run failed tasks.
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   A notebook cell runs successfully but returns no output, even though the engineer expects a DataFrame to be displayed. What is the most likely cause?

   • AThe cluster is stopped
   • BThe cell is in Markdown format
   • CThe notebook is attached to the wrong cluster
   • DThe last expression in the cell did not produce a DataFrame
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A data engineer is analyzing a query profile for a Spark SQL query that joins two tables. The query profile shows that the join is performed as a SortMergeJoin, and the shuffle read size is very large. The engineer wants to reduce the shuffle size. The smaller table is 200 MB, and the larger table is 2 TB. What should they do?

   • AUse a broadcast hint on the smaller table.
   • BEnable adaptive query execution (AQE) to optimize the join.
   • CUse a bucketed table for the join columns.
   • DIncrease the number of shuffle partitions.
   Reveal Solution Discussion Select an answer first