Code Room
On-callMediumoc-g349
Subject CorruptionLevel Senior~35 minCommon in Reliability & on-call interviewsIndustries Technology

Question

A nightly job writes Parquet files to your data lake; downstream Spark/Trino jobs read them for reporting. After a library bump in the writer image at 01:00, last night's partitions *write* fine and a few readers succeed, but most reader jobs fail with 'unsupported/unknown compression codec' or read garbage column values, while older partitions read perfectly. Dashboards: writer job exit code 0, file counts and sizes look normal in object storage. How do you triage, stop producing unreadable partitions, and recover the data?

What a strong answer looks like

Stop the bleeding first (mitigate), then form hypotheses from real signals. Separate root cause from symptom, communicate status as you go, and close with what prevents a repeat.

Diagram & narrate the incident
Loading whiteboard…
Run or narrate your approach, then ask the coach.