Lineage-Based Recovery: Rebuild, Don't Re-Read

When an executor dies mid-job, it takes its partitions of in-flight data with it. A naive system would have to start over, because that data is gone. Spark does not, and the reason is that it never treated those partitions as precious irreplaceable data in the first place. It treated them as the output of a known recipe. The lineage is that recipe, recorded per partition, and recovery is just running the recipe again for the partitions that were lost. Say a partition was produced by reading a chunk of the source, filtering it, and mapping over it. The lineage records just that: this partition came from that source chunk via those narrow operations. When the executor holding it dies, Spark looks up the lineage, sees the recipe, re-reads that one source chunk, and re-applies the filter and m

About This Interactive Section

This section is part of the Lineage as Fault Tolerance lesson on DataDriven, a free data engineering interview prep platform. Each section includes explanations, worked examples, and hands-on code challenges that execute in real time. SQL queries run against a live PostgreSQL database. Python runs in a sandboxed Docker container. Data modeling problems validate against interactive schema canvases. All content is framed around what data engineering interviewers actually test at companies like Meta, Google, Amazon, Netflix, Stripe, and Databricks.

How DataDriven Lessons Work

DataDriven combines four interview rounds (SQL, Python, Data Modeling, Pipeline Architecture) with adaptive difficulty and spaced repetition. Easy problems get harder as you improve. Weak concepts resurface until you master them. Your readiness score tracks progress across every topic interviewers test. Every lesson section ends with problems you solve by writing and running real code, not by picking multiple-choice answers.