1. Q1. A nightly integrity check on a deduplicating repository has reported success every night for six months. What does that establish about the stored data?

   That the repository structure and metadata are self-consistent, and nothing about the stored bytes That every stored object was read and compared against its recorded hash That the most recent recovery point is restorable That no object has been deleted since the last full verification

2. Q2. Ten bytes are overwritten in the middle of an existing repository data file. Which observation would reveal the damage without reading and re-hashing the contents?

   The file size no longer matches the size recorded in the index The directory listing shows the file as truncated None of them; only re-reading the data against a recorded hash detects it The filesystem reports a checksum error on the inode

3. Q3. A service is backed up nightly at 01:00 with no continuous mechanism. It fails at 20:00. What is the honest RPO statement for this architecture?

   Fifteen minutes, because the backup completes quickly Nineteen hours, because that is the elapsed time since the last run One hour, because the schedule can be tightened Up to twenty-four hours, worst case immediately before the next scheduled run

4. Q4. A team quotes a four-hour RTO. Their only offsite copy is in an archive tier whose documented retrieval latency is measured in hours before any transfer begins. What is the correct characterisation?

   A demanding but achievable objective that needs a faster network An architecture mismatch: the retrieval stage alone can exceed the objective An acceptable objective, since retrieval and restore run concurrently A monitoring problem, because the retrieval time is not currently measured

5. Q5. A file-level copy of a running database directory is taken under write load. It starts, completes crash recovery, and returns every expected row. What has been demonstrated?

   That the copy is a transaction-consistent image of a single instant That file-level copies of live databases are safe when the database has a write-ahead log That this copy happened to contain the log records it needed, and nothing about consistency That the database was idle during the copy

6. Q6. Production data lives on one storage array with hourly snapshots retained for two weeks. The array is destroyed. What recovery capability do the snapshots provide?

   Two weeks of recovery points, since snapshots are independent objects The most recent snapshot only, because older ones are consolidated Recovery to any hour, provided the array controller is replaced None: the snapshots were allocations inside the storage that was lost

7. Q7. A fixed-size copy-on-write snapshot is taken of a busy volume and the volume is then heavily rewritten. What consumes the snapshot space?

   Writes to the origin volume, which force the previous contents to be preserved Writes to the snapshot, which is why it must be mounted read-only Reads of the snapshot, which materialise blocks on demand The size of the origin volume at the moment the snapshot was taken

8. Q8. An organisation stores its backup repository passphrase in a file on the server being backed up, so the nightly job can run unattended. The site is destroyed and the offsite repository is intact. What is the recovery position?

   Recoverable, because the key material is inside the repository snapshots Recoverable after brute-forcing the passphrase, since the algorithm is known Not recoverable: every byte survives and nothing can decrypt it Recoverable, because the storage provider can decrypt on request

9. Q9. A backup object is protected by a retention mode that an administrator can override by passing a bypass permission. Which threat does that protect against?

   An adversary who has obtained administrative credentials A mistaken deletion by someone not intending to override the protection A storage platform that silently loses objects A lifecycle rule that expires the object early

10. Q10. An estate mirrors production to a second site every night with deletion propagation enabled. An operator deletes a directory at 14:00. What is the state of the mirror at 02:00?

    The directory is retained, because mirrors keep replaced files by default The directory is present but marked stale for manual review The mirror job fails, because the source no longer matches The directory is gone from the mirror, and the job reported success

11. Q11. A cluster-state snapshot is restored after a control-plane loss. A workload that was scaled up after the snapshot returns at its earlier replica count, and a configuration object created after the snapshot is absent. What has happened?

    The restore failed partially and should be repeated The snapshot was corrupted and silently truncated The snapshot predates those changes; this is the RPO being paid The objects were garbage-collected during the restore

12. Q12. A team keeps every application manifest in version control and treats that repository as their cluster backup. The cluster and its storage are lost. What comes back from the manifests?

    Every object and its persistent data, since the manifests declare the volumes Every declared object, with the persistent volumes created empty Nothing, because object identifiers are not stored in version control The persistent data only, since object definitions are regenerated automatically

13. Q13. A repository is damaged and a restore of a listed recovery point reports a non-zero exit status while writing most of the expected files. What is the correct interpretation?

    The restore is usable, because the majority of files were written The restore is incomplete, and files present on disk may still be wrong The exit status refers to the listing operation, not the data The recovery point is intact and the target filesystem is at fault

14. Q14. A recovery has completed and the database accepts connections. Which observation would justify declaring the recovery successful?

    The service reports healthy and the process has been running for an hour The data directory is the expected size and the schema objects exist The log shows recovery reached the configured target without error An invariant recorded before the incident still holds in the recovered data

15. Q15. Which of these are established by a backup job that exits zero? Select all that apply.

    That a process ran to completion on its schedule That the stored data is intact That the encryption key will be reachable at recovery time That the data can be restored within the required time

16. Q16. Two copies of a dataset are held. Which conditions would make them NOT independent? Select all that apply.

    Both are reachable and deletable using one set of credentials Both live on the same storage array One is a continuous mirror of the other with deletion propagation They are held in different cities on different hardware

17. Q17. A retention mode that refuses deletion for a fixed window still leaves which risks open? Select all that apply.

    An attacker writing large volumes of data that cannot be reclaimed Objects written after a compromise being faithfully preserved The retention window expiring before the compromise is detected Silent corruption of the stored objects being repaired automatically

18. Q18. Which items are commonly absent from an estate inventory and stop a recovery when they are? Select all that apply.

    Infrastructure state files describing what was provisioned The backup catalogue or index describing the recovery points Operating system packages available from a public mirror Certificate authority private key material

19. Q19. A restore that writes files to disk and exits non-zero can leave corrupt files whose names and sizes look entirely correct.

    True False

20. Q20. Enabling deletion-refusing retention on a bucket that already holds a year of backups is a configuration change that can be applied at any time.

    True False

21. Q21. A backup mechanism whose repository is append-only necessarily refuses commands that delete recovery points.

    True False

22. Q22. Measuring how long a backup takes gives a useful estimate of how long the corresponding restore will take.

    True False

23. Q23. An estate holds three copies of its data. State the single test that determines whether the copy count means anything, and name the axis of independence that estates most often fail to model.

24. Q24. A ransomware event is confirmed. Explain why the newest recovery point is usually the wrong one to restore, and what has to be established before any recovery point is chosen.

25. Q25. Produce the evidence you would gather, the remediation you would apply, and the rollback you would hold in reserve.

    A service has been backed up nightly for two years to a repository on a NAS, using credentials stored on the application server. The dashboard shows 99.8% job success. The team has never performed a restore. The repository is also synchronised each morning to a second NAS in another building using a mirroring tool with deletion propagation enabled. The encryption passphrase is in a file on the application server so the job can run unattended. A director asks whether the service could be recovered after a ransomware event.

Passing score: 80%. Answers are checked in this browser.