Let's see if you think like a senior data engineer. His Spark job takes 20 minutes every day. Today it took two hours. What's the first thing you will check?

The first step is to examine the job’s runtime metrics and logs. I would look at the Spark UI to see if there were any executor failures, increased GC pauses, or a spike in shuffle read/write. Next, I’d compare the current job’s DAG to the previous run to spot any added stages or changes in partitioning that could increase parallelism overhead. I’d also verify that the input data size hasn’t grown unexpectedly; a larger dataset can easily double runtime. If the data volume is unchanged, I’d review recent code commits for new transformations, joins, or broadcast operations that might be more expensive. Finally, I’d check cluster configuration changes such as executor memory, number of cores, or network settings that could affect throughput. By systematically ruling out data, code, and infrastructure factors, I can pinpoint the root cause of the slowdown.