Practice
79 scenarios · 0 played
Interview rounds
Backend and SRE interview scenarios you play, not read. Something has gone wrong: you ask, act, pick an approach and watch it play out, then compare with a model answer and a scoring rubric.
Mock round
Four to six scenarios in a row, mixing round types, with a summary of rubric scores, strengths, gaps and what to study. It uses the domain and level picked below: any domain, Senior.
Progress by domain
Scenarios
79 of 79 shown
Domain
Role
Level
Round type
- Concurrency & payments · Production incidentCancelled, and still chargedAn order fails at its last step, and the step that should undo the payment fails too. How do you keep the system consistent?Backend · Staff · 14 min · plays out · full rubric
- Concurrency & payments · Design under changing requirementsCaptured, then authorised againUpdates to one order arrive late, twice and out of order. How do you design the write so the row still ends up right?Backend · Senior · 14 min · plays out · full rubric
- Concurrency & payments · Production incidentOne left, two soldTwo customers buy the last item at the same moment and both orders succeed. How do you stop stock being sold twice?Backend · Mid · 10 min · plays out · full rubric
- Concurrency & payments · Design under changing requirementsOne limit, six gatewaysA payments API runs on many gateway instances and promises 100 requests a second per API key. How do you enforce that limit?Platform-SRE · Senior · 14 min · plays out · full rubric
- Concurrency & payments · Production incidentOne order, two chargesA retried payment request charges the customer twice. How do you make sure a payment happens once?Backend · Senior · 12 min · plays out · full rubric
- Concurrency & payments · Production incidentPaid, and still pendingThe provider took the payment, and the order still says it is waiting for it. How do you make sure a paid order is marked as paid?Backend · Senior · 12 min · plays out · full rubric
- Concurrency & payments · Bug Squash (Java)The hit counter that counts lowA Java class counts API calls per key from many request threads, and its totals come out low. Where is the bug?Backend · Mid · 10 min · full rubric
- Concurrency & payments · Bug Squash (Java)The limiter that admits one too manyA fixed-window rate limiter in Java allows 100 requests per key per minute, and request 101 gets through. Where is the bug?Backend · Mid · 10 min · full rubric
- Concurrency & payments · Design under changing requirementsThe one-time code that wasn'tDesign sign-in with a 6-digit code sent by SMS. Then: what does your design do when someone attacks it, or the SMS provider fails?Backend · Senior · 14 min · plays out · full rubric
- Concurrency & payments · Bug Squash (Java)The redirect that dropped the paymentA hand-written Java HTTP helper follows redirects, and a POST arrives at the new address with nothing in it. Where is the bug?Backend · Senior · 12 min · full rubric
- Kafka & events · Production incidentA broker died and took orders with itA broker fails and some acknowledged records are gone. What do you configure so that a broker failure cannot lose an acknowledged write?Platform-SRE · Senior · 14 min · plays out · full rubric
- Kafka & events · Trade-off defenceA polling outbox, or the database log?To publish events reliably from a service with a PostgreSQL database, do you poll an outbox table or capture changes from the database log?Backend · Senior · 13 min · plays out · full rubric
- Kafka & events · Production incidentA producer release breaks its readersAfter a producer release, one consumer of an event fails on every record and another reports odd figures. How should the event be changed safely?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentBack after nine days, and nothing to catch up onA Kafka consumer returns after nine days offline and reports no lag, yet its work for those days is missing. How do you recover it?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentBusy consumers, no progressA consumer group keeps rebalancing and its throughput has collapsed. How do you find out why and make the group stable?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Design under changing requirementsCheckout fails when the order service is downCheckout calls the order service synchronously and fails with it. How do you redesign the hand-off with Kafka so checkout keeps working?Backend · Senior · 15 min · plays out · full rubric
- Kafka & events · Production incidentClosed accounts return after a rebuildA service rebuilds its state from a compacted Kafka topic and finds customers who closed their accounts. How should a deletion be written?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentEvery deploy costs twenty minutes of lagConsumer lag jumps after every deploy of a consumer service and takes twenty minutes to recover. How do you make it small?Backend · Mid · 12 min · plays out · full rubric
- Kafka & events · Capacity designHow many partitions, how many consumers?Size a Kafka topic and its consumer group for a known peak rate. How many partitions and consumers do you need?Backend · Senior · 12 min · plays out · full rubric
- Kafka & events · Production incidentMore consumers, same lagOne partition of a Kafka topic lags while the others are idle, and adding partitions and consumers did not help. What do you do?Backend · Senior · 12 min · plays out · full rubric
- Kafka & events · Concept deep-diveOnce, at least once, or not at allWhat delivery guarantees does Kafka give, where do they stop, and how do you get a result that happens once in a database?Backend · Staff · 14 min · plays out · full rubric
- Kafka & events · Production incidentOne record stops a partitionOne Kafka record fails every time it is processed and the partition behind it stops. What should the consumer do with it?Backend · Mid · 12 min · plays out · full rubric
- Kafka & events · Production incidentOrder saved, event missingA service saves to its database and publishes an event to Kafka as two separate writes, and they drift apart. How do you keep them consistent?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentReserved twice after a restartA Kafka consumer is killed after it processed a record and before it committed the offset. What happens next, and how do you make that safe?Backend · Mid · 10 min · plays out · full rubric
- Kafka & events · Production incidentSaved, refreshed, and the old value is backA user saves a change, refreshes, and sees the old value, because the page reads a model built from events. How do you let users see their own writes?Backend · Senior · 13 min · plays out · full rubric
- Kafka & events · Production incidentThe confirmation emails that arrive lateConsumer lag on a Kafka topic is above its alert threshold. Is it a passing spike or a real problem, and what do you do?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentThe lag climbs while the consumers workA consumer group's lag keeps growing although its consumers are alive. How do you find out what is stuck, and what do you change?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentThe parcel that went back in transitEvents for one entity are handled out of order by a Kafka consumer. How do you keep each entity's events in order?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentThe refund that was paid twiceA Kafka consumer processes the same event twice after a restart. How do you make the second delivery harmless?Backend · Senior · 14 min · plays out · full rubric
- Kafka & events · Production incidentTwo consumers, one partitionTwo consumers in one group acted on the same records at the same time. Who owns the partition, and how do you stop the other one's side effects?Backend · Staff · 15 min · plays out · full rubric
- Databases & caching · On-call judgement03:10, and a replica is 90 seconds behindAn alert wakes you at night because a read replica is 90 seconds behind the primary. Is it worth a page, and what should the alert measure?Platform-SRE · Senior · 12 min · plays out · full rubric
- Databases & caching · Concept deep-diveA cache on four machines loses oneHow do you design a cache that runs on many machines, so that losing or adding a machine does not flood the database?Backend · Staff · 14 min · plays out · full rubric
- Databases & caching · Trade-off defenceA relational database or a wide-column store?How do you choose between a relational database and a wide-column store for a new workload?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Trade-off defenceAdd a cache, or fix the query?A read-heavy endpoint is slow. When does a cache in front of the database help, and when is it needless risk?Backend · Mid · 12 min · plays out · full rubric
- Databases & caching · Production incidentAfter the failover, half the orders go missingThe database primary failed over in the middle of a deploy, and some confirmed orders are now missing. How should a failover work so that every write goes to one primary?Platform-SRE · Senior · 14 min · plays out · full rubric
- Databases & caching · Capacity designHow many connections does the sale need?Size the database connection pools of a service for a known peak. How many connections does each instance need, and how many can the database take?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentNo connections left, and an idle databaseRequests time out waiting for a database connection while the database is almost idle. How do you stop the pool from running out?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentOne cache node is saturated, five are idleOne node of a Redis Cluster is saturated while the others are idle, and reads through it are slow. How do you bring latency back down?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentOne entry expires, the database goes downA burst of cache misses overloads the database at the busiest time of day. How do you stop a missing cache entry from doing that?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Design under changing requirementsRenaming a column while two versions runA column on a busy PostgreSQL table must be renamed while old and new versions of the service run side by side. How do you do it with no failed requests?Backend · Senior · 14 min · plays out · full rubric
- Databases & caching · Production incidentSaved, and the old value is still thereA user saves a change and the next page still shows the old value. How do you make sure users see their own writes?Backend · Mid · 10 min · plays out · full rubric
- Databases & caching · Production incidentSlow at peak, with no deadlock in sightAt peak, some writes wait many seconds in PostgreSQL and no deadlock is reported. How do you find what they wait for and stop it?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentThe database is right, the cache is an hour behindA cached API sometimes serves an old value long after the data was changed. How do you stop the cache from holding an older value than the database?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentThe primary's disk is filling, and no table grewDuring a large backfill the PostgreSQL primary's disk heads towards full. How do you keep the primary up?Platform-SRE · Staff · 14 min · plays out · full rubric
- Databases & caching · Production incidentThe query that got ten times slower overnightA production query is suddenly ten times slower and nothing was deployed. How do you find out why and bring it back?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Production incidentTransfers that abort each otherUnder load, some transactions fail with a deadlock error from PostgreSQL. How do you stop them deadlocking?Backend · Senior · 12 min · plays out · full rubric
- Databases & caching · Concept deep-diveWhy not put an index on every column?A query is slow and someone proposes an index on every column. What does an index cost, and how do you decide which ones to have?Backend · Mid · 12 min · plays out · full rubric
- Java/Spring/JVM · Production incidentEvery endpoint is slow at onceEvery endpoint of a service is slow at the same time while its CPU is idle. How do you contain it?Backend · Senior · 12 min · plays out · full rubric
- Java/Spring/JVM · Production incidentEvery test passed, and the app crashedA change to a REST response breaks some clients after the deploy. How do you change an API without breaking the clients already out there?Backend · Mid · 10 min · plays out · full rubric
- Java/Spring/JVM · Production incidentOut of memory every nightA service runs out of heap and restarts every night. How do you tell a leak from a heap that is too small?Backend · Senior · 12 min · plays out · full rubric
- Java/Spring/JVM · Bug Squash (Java)The transfer that would not roll backA Spring @Transactional method throws, and the database change is not rolled back. Where is the bug?Backend · Senior · 10 min · full rubric
- Java/Spring/JVM · Production incidentUp, idle, and answering nothingA service stops answering while its CPU is close to zero. How do you find out what its request threads are stuck on?Backend · Senior · 12 min · plays out · full rubric
- Kubernetes & deploys · Production incidentA JVM pod that restarts, vanishes or crawlsA JVM service's containers keep restarting at peak. What do you change so that they stop?Platform-SRE · Senior · 14 min · plays out · full rubric
- Kubernetes & deploys · Production incidentRunning for an hour, Ready for none of itThe new pods are Running, never become Ready, and are restarted again and again. What do you change so that they start and take traffic?Backend · Mid · 10 min · plays out · full rubric
- Kubernetes & deploys · Production incidentThe deploy succeeds, and customers see errorsEvery rolling deploy reports success, and every one produces a burst of 5xx. What do you change so that a rollout drops no requests?Backend · Senior · 12 min · plays out · full rubric
- Kubernetes & deploys · Production incidentThe failover that half workedThe primary region fails, traffic is moved to the second region, and things go wrong there. How should a failover between regions be designed?Platform-SRE · Senior · 14 min · plays out · full rubric
- Kubernetes & deploys · Production incidentThe new pods will not stay upAfter a deploy the new pods are in CrashLoopBackOff. How do you find out why the container keeps exiting?Platform-SRE · Mid · 10 min · plays out · full rubric
- Observability · Design under changing requirementsA billion points a minuteDesign a metrics pipeline that ingests billions of points a minute and still answers queries in seconds.Platform-SRE · Staff · 12 min · full rubric
- Observability · Production incidentThe outage nobody was paged forThe monitoring system is down during an outage. How do you design a fallback?Platform-SRE · Senior · 10 min · full rubric
- Observability · Production incidentThe label that tripled the billCardinality exploded after a deploy and costs tripled overnight. How do you find the label and stop it?Platform-SRE · Senior · 12 min · plays out · full rubric
- Observability · On-call judgementThe pager everyone learned to ignoreAlert volume doubled and on-call is ignoring pages. How do you fix it?Backend · Senior · 10 min · plays out · full rubric
- Observability · Production incidentAll green, and the customers cannot payDashboards are green but customers report errors. What did you fail to measure?Backend · Senior · 10 min · full rubric
- Observability · Design under changing requirementsThe slow trace that was never keptWhen would you use tail-based sampling over head-based, and what does each lose?Platform-SRE · Senior · 10 min · plays out · full rubric
- Observability · On-call judgementPages for the CPU, silence for the customersHow do you design alerting that pages on customer impact, not CPU?Backend · Senior · 12 min · plays out · full rubric
- Observability · Production incidentTwenty minutes behindA Kafka-backed ingest pipeline lags by 20 minutes. What do you check, in order?Platform-SRE · Staff · 12 min · plays out · full rubric
- Observability · Production incidentOne tenant, ten times the dataA noisy tenant sends 10x normal data. How do you contain it fairly?Platform-SRE · Staff · 12 min · plays out · full rubric
- Observability · Design under changing requirementsVisibility without touching the codeHow would you use eBPF for visibility without instrumenting code, and what are its limits?Platform-SRE · Staff · 10 min · full rubric
- Observability · Design under changing requirementsThirteen months at full resolutionHow do you design hot, warm and cold retention for cost versus usefulness?Platform-SRE · Senior · 10 min · full rubric
- Observability · Production incident"The service is slow"A service is reported as slow, with no alert and no obvious cause. How do you troubleshoot it?Backend · Senior · 15 min · plays out · full rubric
- Observability · On-call judgementFive engineers, one pager, and nobody is sleepingA team of five is worn out by its on-call rotation. How would you redesign it, and what do you do when you disagree with a call during an incident?Backend · Senior · 14 min · plays out · full rubric
- Observability · Production incidentOne country says it is slow. Every dashboard is greenUsers in one region report slow responses while the global dashboards show nothing. How do you find the cause and what do you do?Platform-SRE · Senior · 14 min · plays out · full rubric
- Observability · Production incidentOne pod at 100% CPU, seven peers at 30%One pod of a service sits at 100% CPU while its peers are fine. What do you check, and in what order?Platform-SRE · Mid · 13 min · plays out · full rubric
- Observability · Production incidentSlow requests that the application cannot explainRequest latency is high and nothing in the application accounts for it. Where do you look next?Platform-SRE · Senior · 15 min · plays out · full rubric
- Observability · Concept deep-diveThe average is fine, and users time outAverage latency looks healthy while users hit timeouts. Why, and what should you measure and alert on?Backend · Senior · 12 min · plays out · full rubric
- Observability · On-call judgementThe budget is gone and the launch is on MondayThe service has spent its 30-day error budget by mid-month, and product wants to ship a large feature. What do you do?Backend · Senior · 12 min · plays out · full rubric
- Observability · Production incidentThe trace ends where the queue beginsA request is slow end to end and its trace stops at a message queue. How do you get one trace across the queue?Backend · Senior · 14 min · plays out · full rubric
- Observability · Production incidentThree alerts, three services, four minutesSeveral alerts fire at once across three services. How do you run the first ten minutes?Backend · Senior · 14 min · plays out · full rubric
- Observability · Trade-off defenceThree nines or four?How do you choose an availability SLO for a service: 99.9% or 99.99%?Backend · Senior · 12 min · plays out · full rubric
- Observability · Capacity designTraffic grows 15% a month. When do you run out?A service's peak traffic grows 15% a month. When does it run out of capacity, and what do you provision and alert on?Platform-SRE · Mid · 12 min · plays out · full rubric
Curated rounds
Fixed sets on one theme, played in order.
Invented scenes, checked answers.Every scenario is written for teaching and labelled as such. None is a record of an interview. The numbers in the animations are illustrative, and the behaviour each answer rests on is checked against the official documentation linked under it.