Spark jobs rarely fail loudly; they slow down, stall or behave unpredictably. This often leaves engineers staring at dashboards wondering: what actually went wrong?

In this session, you will learn how to approach Spark performance issues like a detective, using the Spark UI and Spark History Server as your primary investigative tools. Rather than touring every tab, we focus on a practical workflow.

Through real-world examples, we will follow the clues: long-running stages, uneven task distributions, high shuffle wait times, and executor imbalances. You will also see how Fabric’s extended history server can accelerate root cause analysis after a job completes.

By the end of this session, you will have a repeatable approach to diagnosing Spark issues in Fabric, understand which signals matter most, and know exactly where to look first when things go wrong.

If you’ve ever asked “why is this Spark job slow?”, this session will give you the tools to answer it confidently.