Exam Prep · SRE Foundation (SREF)

SREF Practice Questions — the Bank

Everything on this page exists to make one thing true before you sit the real SREF: that a single-best-answer question stops feeling like a guessing game and starts feeling like a puzzle with a method. This page isn't itself a 40-question paper — that's what the five full mock exams are for. It's the hub that teaches you how a single-best-answer item is actually constructed, hands you a systematic way to eliminate the options that are built to tempt you, walks three real questions through that process end to end, and then routes you into the three module-grouped practice sets — 64 questions, eight per blueprint module, sat one focused group at a time instead of all forty at once. Read the method here first; the grouped banks are where you drill it until it's reflex.

The anatomy of a single-best-answer item

☺ Like you're 10: Every question has exactly three parts — the question itself, the one right answer, and three wrong answers built on purpose to look tempting.

Every SREF question is built from the same three parts, and knowing the parts by name changes how you read them. The stem is the question itself — one precise ask, however much scenario dressing surrounds it. The key is the single option that answers that exact stem with no qualification needed. The distractors are the other three options, and none of them are random — each one is deliberately built to be attractive to a reader who has a rough, not a precise, grasp of the material. That's the whole game of "single-best-answer": more than one option is usually defensible in isolation, so the exam isn't testing whether you recognize a true SRE fact, it's testing whether you can find the one option that answers this specific stem better than the other three.

The distractor taxonomy — six ways a wrong answer gets built

☺ Like you're 10: Wrong answers aren't random — they come in a handful of repeatable shapes, and once you can name the shape, it stops fooling you.

Distractors aren't arbitrary noise. Across the SREF blueprint's vocabulary — and across most closed-book, single-best-answer exams built the same way — wrong options fall into a small number of repeatable patterns. Learning to recognize the pattern is faster than re-deriving the correct fact from scratch every time, because spotting "that's a reversed-fact distractor" rules out an option in a second, independent of whether you can also state the correct fact confidently.

The elimination method — six steps, in order

☺ Like you're 10: Cover the answers, guess your own first, then cross off anything false, anything that answers a different question, and anything that sounds too extreme — what's left is almost always right.

This is the process the three worked examples below apply, step by step, and it's the same process you should be running — mostly unconsciously, by the time you sit the real exam — on every one of the 64 questions in the grouped banks.

1. Cover the options and answer from memory first. Read only the stem and try to produce the answer in your own words before you look at A through D. This single habit does more to defeat distractors than anything else on this list, because a distractor can only tempt you if you're comparing it against the other options instead of against a fact you already committed to.
2. Read the stem a second time, isolating the exact question. Underline (mentally or literally) the operative word — "best," "first," "not," "primary," "true opposite." A huge share of wrong answers come from correctly understanding the scenario but answering a slightly different question than the one actually asked.
3. Cross out anything outright false. The lowest bar: does the option state something that isn't true about SRE at all? Reversed facts and fabrications usually die here.
4. Cross out anything true but off-target. Harder, and where most of your time should go: is this option accurate, but about a different term, a different scope, or a different question than the stem asked? "Right concept, wrong scope" and "true statement, wrong question" distractors survive step 3 and die here.
5. Cross out absolute or fabricated-sounding language on sight. "Always," "never," "only," or a term you don't actually recognize from the blueprint are each, on their own, a strong reason to deprioritize an option — treat them as a tiebreaker even before you've fully reasoned through the sentence.
6. Between what survives, pick the option that most precisely matches the stem's exact wording — and don't re-open it. If two options both survive steps 3–5, the answer is almost always the one that maps most literally onto the stem's specific noun and verb, not the one that "feels" more like something you've read. Once you've committed with a real reason, changing your answer without a new reason is usually your reflexes second-guessing your reasoning, not correcting it.

Three worked examples — elimination in action

☺ Like you're 10: Three real-feeling questions, worked all the way through out loud — see exactly which trap each wrong answer was, not just which letter was right.

Unlike the explanations under a mock exam question — deliberately brief, because a mock is meant to feel like the real sitting — the walkthroughs below spell out the full elimination reasoning, option by option, the way you should be narrating it in your own head. Attempt each one cold before opening its answer.

Example 1 — SLOs & error budgets

Q. A service's SLO is 99.95% over a 30-day window. At the end of the window, measured attainment comes in at exactly 99.95% — not below it, precisely on it. A product manager asks whether the team can ship a moderately risky feature today. What should the on-call SRE say?

• A. Yes — the service is meeting its SLO, so there's no reason to hold the release.
• B. No — meeting the SLO exactly means the error budget for the window is fully spent, and the release-freeze policy applies.
• C. Yes — SLAs only take effect once a customer files a formal complaint, so nothing blocks the release yet.
• D. No — SLOs are compliance targets set by the legal team, and only legal can authorize an exception.

Walk the elimination

A is the classic "haven't breached yet ≠ have budget left" trap: it sounds reassuring, but sitting exactly at the SLO line means the budget is entirely spent, not merely undisturbed — true statement, wrong question (it answers "did we breach?" when the real question is "do we have budget left?"). C pulls in a real term, SLA, but from the wrong scope entirely — an SLA is the external contract, not the mechanism that gates a release decision — and invents a "only after a complaint" rule that isn't part of any SLA mechanic: right concept, wrong scope, plus a fabrication. D is an absolute-language fabrication — SLOs are internal engineering targets the team itself sets, not a legal compliance instrument, and there's no such authorization process. B survives all three cuts and matches the stem precisely: the answer is B.

Example 2 — anti-fragility

Q. Which statement correctly describes the relationship between "robust" and "antifragile" as Nassim Taleb's framework is commonly applied in SRE?

• A. They're synonyms — both describe a system that actively improves under stress.
• B. Robust is the true opposite of fragile; antifragile is simply a stronger form of robust.
• C. Robust resists a stressor and stays essentially unchanged; antifragile is the true opposite of fragile and actively gains from the stressor.
• D. Antifragile means a system that never fails, regardless of load.

Walk the elimination

A is an outright fabrication — robust and antifragile are explicitly distinct points on the same spectrum, not interchangeable labels for the same behavior. B is the single most common wrong answer this exact question produces: it's the reversed-fact trap, since robust is the neutral middle of the spectrum (a shock does nothing), while antifragile is the true opposite of fragile — treating one as "more of" the other erases the distinction the question is testing. D is absolute language dressed as a definition — no real system is unconditionally invincible "regardless of load," and antifragility describes a direction of change under stress, not a claim of immunity from failure. C is the only option that keeps both terms distinct and correctly identifies antifragile, not robust, as fragile's true opposite: the answer is C.

Example 3 — monitoring & SLIs

Q. A team pages on-call whenever a database's CPU crosses 85%, even on shifts where checkout latency and error rate stay completely flat the entire time. What's the most accurate critique of this alerting rule?

• A. None — CPU is one of the four golden signals, so this is a textbook example of correct alerting.
• B. It's cause-based rather than symptom-based, which risks paging on a condition that never actually affects users and trains responders to start ignoring pages.
• C. It's fine, since USE and RED are the same framework applied to different layers of the same system.
• D. It's incorrect because 85% is too low a threshold; the alert should fire at 95% instead.

Walk the elimination

A is true about a different question — CPU/saturation is indeed part of the golden-signals vocabulary, but the stem isn't asking "is CPU a recognized signal," it's asking whether paging on it here is good design; true statement, wrong question. C is a fabrication dressed as a fact — USE (Utilization, Saturation, Errors) targets one resource and RED (Rate, Errors, Duration) targets a request-driven service end to end; they're related but genuinely distinct frameworks, not the same one relabeled. D accepts the premise that a number is the problem and argues about which number — but the structural issue is alerting on an internal cause instead of a user-facing symptom at all; raising the threshold doesn't fix that, it just delays the same wrong kind of page. B names the actual structural flaw and its real consequence, matching the stem precisely: the answer is B.

The three grouped banks — 64 questions across all eight modules

☺ Like you're 10: Instead of one giant pile of forty mixed-up questions, the practice bank is split into three smaller piles, so a weak spot in one topic shows up fast instead of getting buried in a big score.

Each of the eight blueprint modules gets eight single-best-answer questions of its own — 64 questions total — split across three sets grouped by how the modules naturally cluster, so a single sitting stays focused rather than jumping topic every question the way a full mock deliberately does. Work them in blueprint order the first time through the course; once all three are done, use the mock exams to test the interleaved, all-eight-at-once condition the real sitting actually is.

The grouping isn't arbitrary. The first bank covers the three modules most people find sequential — you can't reason well about error budgets (Module 2) without the SRE framing from Module 1, and toil (Module 3) is the operational discipline error budgets exist to protect. The second bank pairs the two modules about instrumentation and the tooling built on top of it. The third bank groups the three modules that live more in culture and process than in arithmetic — anti-fragility, organizational structure, and how SRE sits alongside the frameworks around it.

How to work the bank

☺ Like you're 10: Cold and timed, one small pile at a time — and when you get one wrong, write down which kind of trap it was, not just the right letter.

Four habits make these 64 questions worth far more than reading them once.

One — go in cold, no notes open. If you've got a tab open to Know It Cold — SREF or the blueprint page while you answer, you're not measuring your recall, you're measuring your ability to search a page — a skill the closed-book exam gives you zero opportunity to use.

Two — narrate the elimination out loud or on paper, even when you're confident. The habit that saves you on exam day is the one you've rehearsed dozens of times beforehand, not the one you suddenly remember to do under pressure for the first time.

Three — when you're wrong, log which distractor type caught you, not just the correct letter. "I fell for a reversed fact on Module 6" is an actionable, specific note. "Got Q14 wrong" tells you nothing to fix next time. A pattern of falling for the same distractor type across multiple modules is itself useful information — it usually means you need to slow down at step 2 specifically.

Four — re-run any set you scored below roughly 80% on, a day later, without re-reading the explanations first. A question you get right the second time for the same reason as the first time is genuinely learned. A question you only get right because you remember which letter you picked last time hasn't taught you anything at all.

That's the method: the anatomy of an item, the six distractor types, a six-step elimination process, and three questions walked through it in full. Go apply it against the three grouped banks — SRE Principles, SLOs & Toil, Monitoring & SRE Tools, and Anti-Fragility & Organizational Impact — then measure yourself against the interleaved, all-eight-at-once condition of a full 40-question mock exam. If a specific term-pair keeps costing you points no matter how carefully you eliminate, that's not an elimination problem anymore — go tighten the underlying definition on the SREF Concept Reference. For the exam-day tactics that sit alongside this method — pacing, flagging, when to guess and move on — see Closed-Book Strategy — No Docs Map, and for a deeper library of near-miss traps by category, Answer Triage — SREF.

Check your answers

1. Stem (the question), key (the one correct option), and distractors (the other three, each built deliberately). Any two of: reversed/opposite fact; right concept, wrong scope; true statement, wrong question; absolute language; plausible-sounding fabrication; reversed order/chronology.
2. ① Cover the options, answer from memory first. ② Re-read the stem, isolate the exact question. ③ Cross out anything outright false. ④ Cross out anything true but off-target (answers a different question or scope). ⑤ Cross out absolute-language or fabricated-sounding options on sight. ⑥ Between what survives, pick the option matching the stem's exact wording, and don't re-open a considered answer without a new reason.
3. SRE Principles, SLOs & Toil covers Modules 1–3 (24 questions); Monitoring & SRE Tools covers Modules 4–5 (16 questions); Anti-Fragility & Organizational Impact covers Modules 6–8 (24 questions). 64 questions total, eight per module across all eight modules.
4. Meeting an SLO exactly means attainment sat precisely on the line — which spends the entire error budget for the window, leaving zero margin. That's functionally identical to having already been slightly under the SLO: the release-freeze policy applies either way, so shipping a risky change now has no budget behind it if anything goes wrong.
5. A mock exam's explanation is deliberately brief — one or two sentences confirming the key — because a mock is meant to feel like the real, time-pressured sitting. This page's worked examples instead spell out the full reasoning for every option, including exactly which distractor type each wrong answer was, because the point here is teaching the method itself, not simulating exam conditions.