Imagine opening an exported order table after a recovery rehearsal. Twelve rows went into the backup. Twelve rows came back. The summary says the counts match.
In this fictional exercise, the expected records have identifiers R01 through R12. The restored table contains R01 through R11, plus a second row labeled R11. R12 is absent.
The row count is twelve. The number of distinct expected identifiers recovered is eleven. A matching count has hidden a missing record behind a repeated one.
We want a recovery demonstration to reach the work you would need to resume. The comforting total at the top of the table is a place to start checking, not the end of the rehearsal.
Keep the expected set
For this made-up case, write three separate results:
Rows restored: 12.
Distinct expected identifiers present: 11 of 12.
Exceptions: R12 missing; R11 repeated.
Keep the expected identifiers in a separate reference before the exercise starts. Otherwise, the restored output becomes its own answer key. It cannot tell you which absent identifier should have been there.
The repeated R11 also needs examination. Its label alone does not establish whether its contents match the other R11 row, whether one version is newer, or which should be retained. Do not silently remove a row to make the counts look tidier. Record the exception and resolve the contents against the reference.
Check what can resume
Now suppose every expected identifier is present once. That passes the identity check in this exercise. It still leaves other questions: are the required fields present, can the intended reader open the result, and does the restored version contain the expected values?
Choose the checks that fit the work you are rehearsing. For an order record, a matching identifier would still leave its amount, date and status to compare. Keep those expected values beside the reference identifier so the next check has an answer key too.
Our earlier version-check lesson follows another part of that record: which state of the work was used?
The version-check lesson
For one low-impact recovery rehearsal, write the expected identifiers and one essential field before looking at the output. Review the returned identities and that field against the reference, then keep missing, repeated or changed entries visible in the result.
The Rebuild Editorial


