ISC · Information Systems and Data Management · 6 practice questions
When to Trigger Failover: RTO and RPO
Trigger failover when a reliable estimate shows primary recovery will miss the RTO, not when the deadline arrives. Below: follow an outage from the first recovery estimate to the failover decision, checking data loss and cutover time.
Try one first
Hint
Focus on exactly what event triggers the policy: is it elapsed downtime, or the expectation that downtime will exceed a stated limit?
Answer A. The policy is triggered when the outage is expected to exceed 2 hours, not when 2 hours have already elapsed. Once IT reasonably estimates a 3-hour restoration time, the activation criterion is met and the disaster recovery site should be activated promptly to minimize total downtime and preserve availability.
Why not B: This is tempting because the 2-hour figure may be read as an elapsed-time threshold. However, the policy activates on the expectation that downtime will exceed 2 hours; waiting until 2 hours have passed would delay recovery unnecessarily.
Why not C: This distractor confuses availability with data integrity. The policy is about expected outage duration (availability), so data loss or corruption is not required to trigger activation.
Why not D: Some may assume disaster recovery is only for major site disasters, but availability policies often require failover for critical-system outages that meet the downtime threshold. The vignette gives no limitation to full-site events.
Step by step
- Set the recovery deadline
A retailer's order-entry platform stops processing at 8:00 a.m.; its RTO is four hours and its RPO is 15 minutes. Service must therefore be restored by noon.
- Read the initial estimate
At 8:25 a.m., operations expects primary recovery by 11:00 a.m. That estimate is within the RTO, so it does not trigger the plan's expectation-based failover rule.
- Recognize the decision-changing update
At 9:10 a.m., operations revises primary recovery to 1:30 p.m., beyond the noon deadline. This update triggers failover now, even though four hours have not elapsed.
- Check the recovery copy
The alternate environment has intact replicated data through 7:55 a.m., five minutes before processing stopped. That recovery point satisfies the 15-minute RPO.
- Include cutover time and act
The alternate environment supports all critical functions and takes 35 minutes to activate. Initiating failover at 9:10 a.m. restores service at 9:45 a.m.; waiting until noon would miss the RTO.
Key points
- Remember: the RTO is a recovery deadline, not a waiting period before failover starts.
- A revised recovery estimate can trigger failover even when the earlier estimate supported continued local recovery.
- Meeting the RPO does not remove the need to meet the RTO.
How the exam traps you
- Wait until noon because the four-hour RTO has not expired. Act at 9:10 a.m., when the revised estimate first predicts recovery beyond noon.
- Keep waiting because the recovery copy meets the RPO. RPO limits data loss. It does not extend the downtime allowed by the RTO.
- Treat initiating failover as restoring service. Include the 35-minute cutover: starting at 9:10 a.m. restores service at 9:45 a.m.
Question 2
Hint
Compare the SLA RPO (5 minutes) to the observed asynchronous replication lag (10-20 minutes) and weigh the risk of propagating recently corrupted writes against the DBAs' ability to recover within the RTO.
Answer D. Because replication is asynchronous with a 10-20 minute lag (exceeding the 5-minute RPO), promoting the DR copy risks unacceptable data loss and may propagate corrupted recent writes. The immediate priority is to contain further corruption by isolating the primary storage and suspending new writes, then perform point-in-time recovery on a clean copy, an action DBAs can complete within the 1-hour RTO. A controlled manual failover is appropriate only if the primary cannot be recovered.
Why not A: Tempting because it restores availability quickly, but incorrect: the DR copy lags 10-20 minutes (greater than the 5-minute RPO) and may already contain corrupted recent writes; automatic failover risks propagating corruption and unacceptable data loss.
Why not B: Tempting because it yields a clean dataset on DR, but incorrect: restoring the nightly backup sacrifices all transactions since that backup (far exceeding the 5-minute RPO) and likely causes more downtime than performing the DBAs' point-in-time recovery, which is feasible within the RTO.
Why not C: Tempting because it avoids immediate disruption, but incorrect: continuing to accept writes risks further corruption and potential replication of corrupted data to DR; immediate containment and remediation are required to protect data integrity and meet recovery objectives.
Question 3
Hint
Focus on what the availability commitment is actually promising and which action best meets the stated recovery objectives under the facts given.
Answer B. Availability obligations prioritize restoring service within the stated RTO and minimizing data loss relative to the RPO. Here the tested standby is approved for production and replication lag (4 minutes) is well within the 15-minute RPO, so prompt failover best meets the availability objective; forensic evidence (logs) can and should be preserved while restoring service.
Why not A: Tempting because candidates often prioritize full root-cause diagnosis. It is wrong because waiting would unnecessarily extend downtime when a tested standby meets the RTO/RPO and there is no sign that failover would spread a security or integrity issue.
Why not C: This sounds cautious but is inferior here: the warm standby is far more current (4 minutes lag) and is already tested for production, so backup restore would likely take longer and increase data loss relative to failing over to the standby.
Why not D: Communication is important, but the stem explicitly removes any requirement to notify customers before recovery. Prioritizing notification over restoration would violate the availability commitment and likely increase outage duration.
Question 4
Hint
Separate transient/component alerts from a confirmed site outage; weigh asynchronous RPO, the disabled automated‑failover policy during business hours, and the 60‑minute RTO when choosing when to cut over.
Answer D. Option D best balances meeting the 60‑minute RTO with the risks inherent in asynchronous replication and the business‑hours setting that disables automated failover. Requiring confirmation from independent monitors and short hands‑on recovery attempts ensures the outage is a true site failure before performing a manual cutover, limiting unnecessary failovers and the risk of split‑brain or data loss. This approach preserves human control for a high‑risk decision while still enabling failover early enough to meet the RTO.
Why not A: Tempting because it prioritizes speed, but a first alert may indicate a transient or component problem rather than a site outage. Immediate failover with asynchronous replication increases the risk of unnecessary data loss and split‑brain, and it bypasses the business‑hours safeguard intended to require human confirmation for site‑level cutovers.
Why not B: Tempting if one misinterprets the RTO as the trigger to act, but the RTO is the target for completing recovery, not the point to start failover. Waiting until the RTO expires can unnecessarily prolong downtime and risk breaching the SLA when an earlier, confirmed manual failover would better balance availability and data integrity.
Why not C: Tempting because automation can be fast, but re‑enabling automated failover during business hours removes the intentional safeguard and increases the chance of split‑brain and data loss with asynchronous replication. Changing automation policy mid‑incident is risky; human confirmation and controlled manual failover are safer for site‑level failures.
Question 5
Hint
Compare the RPO to replication lag (synchronous vs asynchronous) and compare each option's estimated recovery time to the RTO.
Answer A. Promoting the local hot-standby preserves committed transactions because it uses synchronous replication, satisfying the 1-minute RPO. The estimated promotion and redirect (~30 minutes) meets the 4-hour RTO. By contrast, the asynchronous remote site can lag up to 15 minutes (risking RPO breach) and restoring from backups (12 hours) would exceed the RTO.
Why not B: Restoring from offsite backups could recover the original primary, yet the estimated 12-hour restore time far exceeds the 4-hour RTO, so it would not meet the availability target.
Why not C: This is tempting because the remote DR is available, but asynchronous replication can lag up to 15 minutes and therefore could lose committed transactions, violating the 1-minute RPO requirement.
Why not D: While thorough integrity checks are prudent, delaying the redirect for an estimated 6 hours would exceed the 4-hour RTO; it's better to restore service immediately via the synchronous standby and perform deeper checks in parallel.
Question 6
Hint
Work backward from the 60-minute restoration objective by subtracting the time required to complete failover and validation.
Answer D. The runbook sets a decision threshold of 35 minutes from disruption start because failover itself requires 25 minutes and the overall availability objective is 60 minutes. At 10:08, the estimate implied total restoration by 10:28, so the threshold had not yet been crossed. At 10:12, the revised estimate implied restoration no earlier than 10:52, which is beyond the 35-minute trigger point, so failover should be initiated then rather than waiting for an actual breach.
Why not A: This is tempting because candidates often treat failover as the automatic response to any outage. It is wrong because the scenario gives a specific decision rule: trigger failover when projected primary recovery exceeds the threshold, not immediately upon every disruption. Immediate failover could be unnecessary if primary recovery is still likely within the allowed time.
Why not B: This distractor draws attention to the 25-minute failover duration, which seems operationally important. But the 25 minutes is not the trigger by itself; it must be evaluated against the 60-minute restoration objective, producing a 35-minute decision point. At 10:08, the projected total outage was still only 28 minutes, so management had not yet reached the runbook threshold.
Why not C: This is tempting for candidates who confuse a recovery objective with a post-event measurement point. It is wrong because availability decisions are made prospectively; waiting until 11:00 would guarantee there is no remaining time to execute a 25-minute failover and still meet the objective. The trigger is based on projected inability to recover in time, not on an already completed failure to do so.
Common questions
Should failover start before the RTO expires?
Yes, when the policy uses expected recovery time as its trigger. Here, the 9:10 a.m. estimate shows primary recovery will miss noon, so management should act then.
Does meeting the RPO mean failover can wait?
No. The recovery copy is only five minutes behind the outage, satisfying the 15-minute RPO, but projected primary recovery still exceeds the four-hour RTO.
Does failover time count toward the RTO?
Yes. The RTO covers the outage through service restoration, not just the time before failover begins. Here, the 35-minute cutover ends at 9:45 a.m., before the noon deadline.
Watch it solved
A full CPA ISC task-based simulation on System availability, worked step by step.
System Availability & RTO/RPO | CPA ISC Simulation (TBS) on YouTube
Practice ISC like the real exam
The free ChatCPA simulator: real exam layout, timed testlets, starting with a question on this topic. No account needed to start.
Open the free simulator →More on System availability
- Choose Disaster Recovery Architecture by RTO and RPO
- System Availability: Redundancy and Automatic Failover
- How to Calculate SLA Uptime at 99.95% Availability
- RTO vs RPO: Synchronous vs Asynchronous Replication
- System Availability: Why Backups Do Not Prove Recovery
- Availability vs. Security, Confidentiality, and Integrity
- Availability Calculation: MTBF, MTTR, and Redundancy
Questions from the ChatCPA bank of 17,658 CPA exam questions, each with a written reason for every wrong answer. ChatCPA is built by Nicholas Miller, CPA (Oregon #14907). How these pages are made. Spot an error? Tell us.