Skip to content

I. Learning quiz

Read the examples first.

A payment service is down for an hour after an engineer runs a command against production instead of staging. The outage is above the team's severity line, and you write the postmortem. Which of the following findings and actions follow the blameless approach?

Select all that apply.

Show answer
  • Correct: Ask why one command could reach production with no guard, and add a confirmation step.. Right. P80 looks for the cause in the system. A missing guard is a system weakness that a fix can remove.
  • Incorrect: Record the root cause as "operator error" and close the report.. Wrong. "Operator error" as the final root cause is the violation named in P80. It names a person and changes nothing in the system.
  • Correct: List the system actions with owners, such as separate credentials and a clearer prompt.. Right. The P80 check asks for a postmortem with system-level actions for each failure above the severity line.
  • Incorrect: Name the engineer in the report so that colleagues can learn from the mistake.. Wrong. P80 says blame makes people hide information. Naming the person to blame is what the rule asks you not to do.
  • Incorrect: Skip the postmortem, because the fix took only ten minutes.. Wrong. P80 applies to each failure above a set severity. The time to fix is not the test, and the outage lasted an hour.
Your team reviews four postmortems from the last quarter. All four failures were above the severity line. Which of them meet the P80 check?

Select all that apply.

Show answer
  • Incorrect: A deploy failure whose report ends with "be more careful" as the only action.. Does not meet P80. "Be more careful" is not a system-level action. The change process (P64) was never examined.
  • Correct: A database outage whose report lists a missing alert and an unreviewed config change, each with an owner.. Meets P80. The causes sit in the system, and the actions fix the monitoring and the change review (P64).
  • Incorrect: A cache failure that was discussed in chat, with no written postmortem.. Does not meet P80. The check needs a postmortem for each failure above the severity line, with actions on record.
  • Correct: A login outage whose report names no person as the cause and sets two tracked system changes.. Meets P80. It finds the root cause in the system and the actions are system-level.
  • Incorrect: A queue overflow whose report blames the on-call engineer for a slow reaction.. Does not meet P80. It stops at a person and does not ask why the system made the slow reaction likely.
A team holds a retrospective every sprint. Three sprints in a row, a release slips because of the same flaky deployment step. Which of the following steps apply Plan-Do-Check-Act?

Select all that apply.

Show answer
  • Correct: Open a ticket with an owner for each retrospective action and review the open ones at the next retrospective.. Right. The P81 check is that actions from earlier cycles are tracked until they are completed.
  • Incorrect: Hold retrospectives only after a release has failed badly.. Wrong. P81 applies always, not only after a failure. A cycle that runs only on failure is not a cycle. P80 is the rule that waits for a failure.
  • Correct: Check whether last sprint's change worked, then keep it or adjust it.. Right. Check and Act close the P81 cycle. Without them the team plans and does, but does not learn.
  • Incorrect: Close each action when the retrospective ends so the list stays short.. Wrong. Closing an action before the work is done breaks the P81 check. Actions stay open until completed.
  • Correct: Change the process, for example with an automated deployment test, instead of listing the same finding again.. Right. P81 names the same failure happening again with no process change as its violation.
  • Incorrect: Blame the engineer who owns the deployment step and move on.. Wrong. P80 looks for the cause in the system, and blame does not feed an improvement cycle (P81).
A department asks how it can show that it runs Plan-Do-Check-Act properly. It lists what it has in place. Which of the following items meet the P81 check?

Select all that apply.

Show answer
  • Correct: Every retrospective action is a ticket with an owner, and open tickets are visible to the whole team.. Meets P81. Actions are tracked, and the check asks for tracking until they are completed.
  • Incorrect: Retrospective notes live on a wiki page that nobody reopens.. Does not meet P81. Notes that are never revisited do not track actions. Monitoring (P70) can supply the data for the Check step, but it does not replace tracking.
  • Incorrect: Open actions are reset at the start of each quarter and unfinished ones are dropped.. Does not meet P81. Dropping unfinished actions ends tracking before they are completed.
  • Correct: Each retrospective opens with the actions from earlier cycles that are still not done.. Meets P81. Earlier actions stay in view until they are completed.
  • Incorrect: The team writes the same three actions in every retrospective, and none is ever closed.. Does not meet P81. Listing is not completing. The check asks for actions to be tracked until they are completed, and none is.
An integration test fails about one run in twenty. Developers rerun the pipeline until it passes. Nobody has been harmed. What should the team do?

Select all that apply.

Show answer
  • Incorrect: Rerun and merge, because no harm means nothing to investigate.. Wrong. P82 applies "even if nobody was harmed". Ignored warning signs come before major accidents.
  • Correct: Open a ticket for the flaky test and decide: fix it, or accept it with a written reason.. Right. The P82 check is a ticket for each anomaly with a decision.
  • Incorrect: Add an automatic retry and mute the alert for good.. Wrong. Silencing an alert permanently without an investigation is the violation named in P82.
  • Incorrect: Wait until a flake blocks a release, then run a postmortem.. Wrong. P80 starts when a failure has happened. P82 asks you to act on the anomaly before it becomes a failure.
  • Incorrect: Wait for a developer to complain, since the test is only a test.. Wrong. P82 asks you to record and investigate the anomaly. It does not wait for a complaint.
A team reviews how it closes anomaly tickets. Which of the following closures meet the P82 check, a decision on each anomaly?

Select all that apply.

Show answer
  • Correct: Closed: sensor glitch only on firmware 2.1, accepted until the Q3 upgrade, review date set.. Meets P82. The anomaly is accepted with a written reason, which the check allows.
  • Incorrect: Closed: won't fix, known issue.. Does not meet P82. Accepting an anomaly needs a written reason, and this one gives none.
  • Incorrect: An engineer calls an alert "noise", and it is silenced for good.. Does not meet P82. Silencing without an investigation is the violation. A noisy alert also fails the actionable-alerts rule (P75), but the first step is to investigate.
  • Correct: Fixed: clock skew between two hosts caused the odd timestamps, patch merged.. Meets P82. The anomaly was investigated and fixed, which is the other allowed decision.
  • Incorrect: The anomaly appears in a chat message and has no ticket.. Does not meet P82. The check needs a ticket for each anomaly.
  • Correct: Alert tuned after tests showed it fires on healthy batch jobs, with the finding in the ticket.. Meets P82. The team investigated first, then decided. This also serves P75, because the alert becomes actionable.
A nightly job takes three times longer than usual for a week. No user has complained and every run still finishes. Which of the following responses fit the taxonomy?

Select all that apply.

Show answer
  • Correct: Open an anomaly ticket the same day.. Right. P82 says to record the anomaly when the system behaves differently from what was expected, even if nobody was harmed.
  • Incorrect: Hold a full blameless postmortem, because every anomaly is a failure.. Wrong. P80 applies after a failure has happened. This is an anomaly, and P82 handles it with a ticket and a decision.
  • Incorrect: Treat the slowdown as normal, because no user complained.. Wrong. P82 says to investigate before you accept it as normal. Lack of complaints is not an investigation.
  • Incorrect: Raise the alert threshold so it stops firing.. Wrong. Changing the threshold to hide the signal is close to the violation in P82, which is silencing an alert without an investigation.
  • Correct: End the ticket with a decision, fix or accept with a written reason.. Right. This is the P82 check. Every anomaly ends with a decision.
A service in a production cluster crash-loops. Users see slow pages, and the team estimates it will be hours before real harm. The on-call engineer wants to restart the pods. Which of the following actions apply P83?

Select all that apply.

Show answer
  • Correct: Capture logs, a heap dump and a disk snapshot before the restart.. Right. P83 says to document and preserve the as-found state before you repair or restart anything.
  • Incorrect: Restart first to restore service and collect evidence later.. Wrong. A repair destroys the evidence needed to find the cause. This is the violation named in P83.
  • Correct: Record who captured the evidence and when.. Right. P83 asks for a record of who handled the evidence, the chain of custody.
  • Incorrect: Roll back the last deploy at once, then try to reproduce the fault in staging later.. Wrong. A rollback is a repair. It removes the as-found state, and a staging copy may not show the same fault. Blameless analysis (P80) needs the evidence.
  • Correct: List the captured evidence in the incident record before the first repair step.. Right. This is the P83 check. The record shows the evidence captured before the first repair action.
  • Incorrect: Delete old pods to free disk space, then take the snapshot.. Wrong. Deleting pods destroys evidence before it is captured. P83 asks you to preserve the as-found state before you change anything.
A control service fails. The temperature of a process rises, and harm is five minutes away. A full manual snapshot would take twenty minutes. Which of the following actions fit the taxonomy?

Select all that apply.

Show answer
  • Correct: Restart the service without waiting for a manual snapshot.. Right. P73 asks that the response is faster than the time until harm. Here that outranks a slow manual capture, as the conflict between P83 and P73 resolves.
  • Incorrect: Spend twenty minutes on a full manual snapshot before any restart.. Wrong. The capture would take longer than the time until harm, which breaks P73.
  • Correct: Run automatic dumps and log shipping that finish in seconds before each restart.. Right. Fast automatic capture serves P83 without costing time against P73.
  • Incorrect: Skip all evidence because harm is imminent.. Wrong. Even when you restore service first, you keep whatever the automation captured. P83 still applies as far as time allows.
  • Correct: Afterwards, keep the automatic dumps and record who handled them in the incident record.. Right. P83 asks for the evidence to be documented and preserved, with a chain of custody.
A dismissed employee files a complaint with the company. The chat tool deletes messages after 30 days, and a related channel is 25 days old. Which of the following actions apply the legal-hold form of P83?

Select all that apply.

Show answer
  • Correct: Suspend routine deletion for the related mailboxes and channels now.. Right. The management form of P83 is to stop routine deletion when a dispute is reasonably anticipated.
  • Incorrect: Wait for a formal letter from a lawyer before changing anything.. Wrong. P83 applies when litigation or an investigation is reasonably anticipated. The hold does not wait for a letter.
  • Incorrect: Keep all company data forever so that nothing is ever lost.. Wrong. The hold covers the relevant records. P83 does not ask for unlimited retention.
  • Correct: Record who exported the preserved records and when.. Right. The legal hold preserves the records, and the chain of custody from P83, a record of who handled the evidence, keeps them trustworthy.
  • Incorrect: Let the manager involved tidy the channel first, to keep it professional.. Wrong. Changing the records before they are preserved destroys the as-found state that P83 protects.
A company reorganizes its support and billing teams. A month later, handoffs fail and customers wait days for refunds. The company plans a review. Which of the following choices apply the after-action review form of P80?

Select all that apply.

Show answer
  • Correct: Hold the review with everyone affected: what happened, why, and how to improve.. Right. This is the after-action review form of P80, run for an event across the teams it touched.
  • Incorrect: Ask each team lead to write down who caused the delays.. Wrong. P80 looks for the cause in the system. Blame makes people hide information, and it works against psychological safety (P5).
  • Incorrect: Let only senior managers meet and send out a summary.. Wrong. The people affected hold the facts about what happened. Leaving them out weakens the analysis, and P5 asks that they can speak openly.
  • Correct: Finish with process changes that have owners, such as a named owner for each handoff.. Right. The aim of P80 is system-level actions. The review ends in changes to the process, not in a verdict on people.
  • Incorrect: Skip the review, because the reorganization is finished.. Wrong. P80 applies after a failure. The handoff failures happened, so the review is due.
Three projects in a row finish about two weeks late, each with a different stated reason. None caused serious harm. Which of the following responses apply the management form of P82?

Select all that apply.

Show answer
  • Incorrect: Judge each project on its own and call it bad luck.. Wrong. P82 asks you to look at deviations even when nothing went wrong. Treating each as an accident hides the pattern.
  • Correct: Record the three misses as one deviation and investigate them as a pattern.. Right. P82 in management means weak signals are recorded and investigated before they become normal.
  • Incorrect: Add two weeks to every plan and treat the slip as normal.. Wrong. P82 says to investigate before you accept the deviation as normal. Padding plans accepts it without a reason.
  • Correct: Change the planning method once the investigation finds the cause, and track the change.. Right. An investigated anomaly ends in a decision (P82), and tracking the change to completion is the P81 cycle.
  • Incorrect: Wait for a project to fail badly before you look into it.. Wrong. P82 acts before harm. Ignored warning signs come before major accidents.