WAL Recovery Checkpoint

Reason through crash snapshots and distinguish redo from restartable undo.


Get every question correct to pass this checkpoint. You can review feedback, revise your answers, and submit again.

1. A loser update at LSN 12 is already present on a page with pageLSN=12. During the one-pass exercise:

Select one answer.

2. A loser writes one byte 4→7 at LSN 5, then 7→9 at LSN 6. Which undo sequence restores the original byte?

Select one answer.

3. P has pageLSN=20 and Q has pageLSN=3. A committed update at LSN 10 targets Q. Redo should:

Select one answer.

4. A committed transaction changed a page that was never flushed. What makes recovery possible under no-force?

Select one answer.

5. T2 aborts, T3 later commits a new value on the same byte, and another crash occurs. Why is blindly undoing T2 again unsafe?

Select one answer.

6. Which claim accurately describes the standalone Recover exercise?

Select one answer.

Join Discord