Exam Prep · SRE Foundation (SREF)

Closed-Book Strategy — No Docs Map

Every other page on this course's SREF shelf assumes the material eventually ends up in your head — this page is about how it actually gets there, because the SREF gives you nowhere else to put it. The CNCF's CNPE, covered by its own Docs Map elsewhere on this platform, is hands-on and narrowly open-book: kubernetes.io/docs, the blog, per-task Quick Reference links, and kubectl explain against whatever's installed in the cluster in front of you. The SREF has none of that. No browser, no terminal, no man page, no cluster to interrogate — a sixty-minute, forty-question, closed-book sitting where the only reference material in the room is whatever you've already memorized. This page is the deliberate opposite of a docs map: instead of telling you where to look things up, it's the strategy for building a working memory system so there's nothing left to look up in the first place — spaced repetition against this course's own flashcards, the active-recall habits that matter far less on a hands-on exam than they do here, and the definition-drilling routine that beats a single-best-answer distractor built around a term you almost, but not quite, know cold.

Zero documentation, not narrow documentation

☺ Like you're 10: CNPE's allowed list is short, not blank; SREF's list is blank, full stop — that's a difference in kind, not just in degree.

It's tempting to read "closed-book" and assume it means something like the CNPE's own restriction — a short allowlist instead of the open internet. It doesn't. The CNPE's own Docs Map establishes that its closed set is still a real, usable set: kubernetes.io/docs and its translations, the Kubernetes blog, whatever task-specific page the exam links from its own Quick Reference box, and documentation installed locally on the exam machine. A CNPE candidate who blanks on the exact field name for a Rollout's canary steps can still recover it — read the Quick Reference link the task handed them, or run kubectl explain rollout.spec.strategy.canary --recursive against a cluster that has the CRD's full schema sitting inside it, permitted, any time. The SREF gives you none of that. There's no browser tab to open, no installed CRD to query, no task-linked reference page — because there's no cluster, no terminal, and no task in that sense at all, just forty multiple-choice questions and whatever's already in your head when the sixty-minute clock starts. "Closed-book" on the SREF isn't a narrower version of open-book. It's the complete absence of the category.

What zero documentation removes from your options

☺ Like you're 10: On a hands-on exam, being unsure costs you thirty seconds while you check. On this one, being unsure costs you the question — there's no "check" button.

The practical consequence is blunter than it sounds. On the CNPE, a candidate who's 70% sure of a field name has a real move available: open the doc, confirm it, move on, and the uncertainty cost thirty seconds instead of the task. On the SREF, being 70% sure of a definition and being 0% sure of it produce the exact same outcome if the 70% happens to be wrong — there's no intermediate step, no confirmation, no partial credit for "I was in the right neighborhood." A single-best-answer question doesn't reward the option that's closest to correct; it rewards the one that's actually correct, full stop, and every other option — including the ones a half-remembered definition would talk you into — scores zero. That's the real argument for the techniques on the rest of this page: not "memorization is a nice-to-have skill for this exam" but "memorization is the entire exam, because there is no second system to catch what memory drops."

Recognition isn't recall — the illusion that sinks closed-book prep

☺ Like you're 10: Humming along to a song and singing it from memory with no music playing are different skills — closed-book exams only test the second one.

The single biggest reason well-prepared candidates underperform on a closed-book exam isn't that they didn't study — it's that they studied in a way that built recognition instead of recall, and mistook the first for the second. Reading a definition of an SLO for the fifth time produces a warm, familiar feeling — the words look right, the concept feels understood — and that feeling is fluency, not retention. Fluency is cheap to build and almost useless on a blank page: it tells you the material is somewhere in your head, not that you can produce it unaided in the ninety seconds a question actually gives you. This is the same trap Know It Cold — SREF is built to fight — it's explicitly written to be recited with the page covered, not re-read a second time — and it's backed by real cognitive science, not just a study tip: the testing effect (retrieval practice measurably outperforms re-reading for long-term retention) and Hermann Ebbinghaus's forgetting curve (unreinforced memories decay fast, then slower, on a predictable schedule) are both well-established findings in memory research, and they point at the same conclusion from different directions — the act of producing an answer, not the act of encountering one, is what actually builds durable recall. Every technique below is a way of forcing production instead of recognition.

Spaced repetition — a manual schedule for this course's 38 flashcards

☺ Like you're 10: Review a fact right before you'd naturally forget it, and each review buys you a longer gap before the next one's needed — that's the whole trick.

Spaced repetition is the direct fix for the forgetting curve above: instead of reviewing everything with equal, arbitrary frequency, you review each individual fact right around when you're about to forget it, which is both more efficient than blanket re-reading and measurably stickier over the long run. The classic implementation is the Leitner system — named after Sebastian Leitner, who described it in the 1970s — where cards live in numbered boxes with increasing review intervals: get a card right, it moves up a box and gets reviewed less often; get it wrong, it drops straight back to Box 1 and gets reviewed again the very next day. Worth being honest about upfront: this course's flashcards deck is a flip-and-search tool, not a scheduling app — it doesn't track which cards you got wrong or space anything automatically. That's not a gap in the deck; it's a gap you're expected to fill yourself, and doing it with nothing but the deck's search box and a plain-text note is entirely sufficient.

Definition drilling — precision over vibes

☺ Like you're 10: "Basically the same idea" isn't good enough here — a single-best-answer question is specifically built to punish "basically."

The SREF's single-best-answer format is unforgiving of approximate definitions in a way a hands-on exam simply isn't — a fuzzy grasp of a term can still get a live cluster into the right state through trial and error, but it can't tell you which of four plausible-sounding sentences is the one the question wants. Know It Cold — SREF already has the core term-pair table this exam leans on hardest — SLI vs. SLO vs. SLA, burning vs. banking, symptom- vs. cause-based alerting — and the SREF Concept Reference has the full definitions behind them; this page is about the drill, not the list. The technique: pick one term, close every reference, and write its full definition from scratch on a blank line — not "kind of describe it," the actual sentence you'd want a grader to see. Then open the concept reference and compare word for word. The gap almost never turns out to be "I didn't know this at all" — it's usually one dropped clause that happens to be exactly the clause a distractor is built around: writing "toil is manual, repetitive work" and dropping automatable, tactical, and devoid of enduring value still describes something real, just not the six-part test the exam is actually checking. That dropped clause goes on the same miss-list as a failed flashcard, and gets re-tested in 48 hours, not "sometime before the exam."

A handful of pairs worth drilling specifically because they sit at the edges of the eight-module blueprint rather than its center, and so are easy to leave under-rehearsed:

Active recall beyond flashcards — blurting, teach-back, and interleaving

☺ Like you're 10: Three more ways to force yourself to produce the answer instead of just recognizing it — a blank-page dump, explaining it to someone imaginary, and mixing topics up instead of studying them one at a time.

Flashcards and definition drilling cover single facts well; three more techniques round out a closed-book routine, each forcing production rather than recognition in a slightly different way:

• The blurt method. Pick one module — say, Reducing Toil — set a timer for ninety seconds, and write down everything you can recall about it on a blank page with no prompts at all: the six-part toil test, the automation ladder's five rungs, the 50% ceiling. Then open the source page and mark what you missed in a different color. The gaps are your study list for that module; the parts you got are confirmed durable and don't need more time.
• Teach-back. Explain a concept out loud, in plain language, to an imaginary colleague who's never heard of it — no jargon allowed as a substitute for understanding. The moment you reach for a term you can't unpack ("it's the burn rate, you know, the burn rate thing") is the moment you've found a gap that recognition-only studying hid from you.
• Mnemonics for anything with a fixed order. Know It Cold's "Please Send Reliable Metrics, Then Automate Our Future" is the ready-made one for the eight blueprint modules — steal it, or build your own for the minutes-per-window table (1,440 / 10,080 / 43,200 / 129,600) if the raw numbers won't stick on their own.
• Interleaving instead of blocking. Studying one module in isolation until it feels solid, then moving to the next, is blocked practice — comfortable, but it doesn't train you to tell two similar concepts apart under pressure, which is exactly what the exam does by mixing modules freely across all 40 questions. Interleaved practice — mixing modules within a single study session — is harder in the moment and produces better exam-day discrimination. The full practice bank and a timed mock exam are naturally interleaved already; use the three focused sets — SLOs & Toil, Monitoring & Tools, Anti-Fragility & Organizational Impact — only for the initial pass on a weak module, then go back to mixed practice.

A closed-book taper — the final seven days

☺ Like you're 10: The last week isn't for learning anything new — it's for finding out what's already slipping, and catching it before the real clock starts.

This isn't a replacement for the SREF study plan's full multi-week schedule — it's the specific closed-book taper for the final stretch, built entirely around forcing recall rather than reviewing content one more time.

That's the strategy: treat the SREF's zero-documentation format as a difference in kind from the CNPE's narrow allowlist, not a smaller version of the same thing, and build every study session around producing an answer rather than recognizing one. Pair this page with Know It Cold — SREF for the exact facts to recite, the Concept Reference for full definitions, Answer Triage — SREF for spotting distractors in context, and the flashcards and the glossary as the raw material every technique above runs against. Then confirm you're actually ready against the practice bank and a timed mock exam before you book the sitting described on the exam guide.

Check your answers

1. The CNPE's allowlist is narrow but real and usable — kubernetes.io/docs and blog, per-task Quick Reference links, and kubectl explain against installed CRDs all remain available during the exam. The SREF permits none of that: no site, no notes, no installed schema to query, nothing at all. It's the complete absence of the category, not a smaller version of it.
2. Recognition is the familiar feeling of "I've seen this and it makes sense," built cheaply by re-reading. Recall is the ability to produce the fact unaided from a blank state, which is what a closed-book exam actually tests. Re-reading mainly builds recognition; it does not reliably build recall.
3. A card answered correctly moves up one box and is reviewed less often (2 days → 4 days → 7 days → 14 days). A card answered incorrectly drops straight back to Box 1 and is reviewed again the very next day, regardless of which box it was in before.
4. MTTD = mean time to detect (failure start → alert fires). MTTA = mean time to acknowledge (page fires → a human accepts it). MTTR is ambiguous by itself (repair, recovery, or resolve depending on the source) — read the surrounding sentence. Availability is the probability the system responds successfully right now; durability is the probability data isn't lost over a long horizon even through temporary unavailability.
5. Blocked practice studies one module in isolation until it feels solid before moving to the next; interleaved practice mixes modules within the same session. The SREF scatters questions from all eight modules across the 40 in no predictable order, so interleaved drilling — via the full practice bank and timed mock exams — trains the actual skill being tested: telling similar concepts apart on demand, not just recognizing each one in its own isolated block.