A bad model is live and the previous artifact is compatible with a new feature schema. How do you roll back?
PICTURE THIS: HOW TO EXPLAIN IT
Simple meaning
If schemas match, move the alias and let Pods reload or roll the Deployment to the old image.
WHY — Rollback instead of guessing?
Why interviewers care about Rollback:
who only read docs from people who shipped.
and tied to MLOps work.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1If schemas match, move
the alias and let Pods reload or roll the Deployment to the old image.
- 2If the new code
cannot load old weights, roll back both code and model as a unit.
- 3Feature flags that keep
dual schemas during migration make this less painful.
- 4Give an example
One tiny concrete case you can say aloud.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
If schemas match, move the alias and let Pods reload or roll the Deployment to the old image. If the new code cannot load old weights, roll back both code and model as a unit.