A Database in Two Seconds: Who Lets the Coding Agent Provision PostgreSQL?
Choose a model for isolated, short-lived test databases and define when a coding agent may request new infrastructure without bypassing platform policy.
Try it →Architecture exercises
Architecture exercises
Start with the newest exercise or choose one of the recent decisions below. Use filters and topic groups when you want a specific area.
Practice library
Recent exercises
01A Database in Two Seconds: Who Lets the Coding Agent Provision PostgreSQL?Data and migrations · Intermediate · Try →02UNKNOWN transaction: who is responsible for closing the case?Integration and APIs · Intermediate · Try →03Every health check is green. Why can't users complete the journey?Resilience and recovery · Intermediate · Try →04The provider delivers SaaS. Do you need to know its architecture?Analysis and decisions · Intermediate · Try →05New optional API field. Can the application still read the response?Integration and APIs · Intermediate · Try →06Should a notification failure block an order?Integration and APIs · Easy · Try →07New tablets, the same app. What should you check?Analysis and decisions · Easy · Try →08No response. Was the booking created?Integration and APIs · Intermediate · Try →
Browse all exercises and filters ↓Matching exercises: 16
No matching articles. Try another term or area.
Data and migrations (3)
Two application versions, one schema. How do you deploy without downtime?
How do you split this change into safe stages? When may writes to the old field stop, when may the field be removed, and how do you test rollback?
Try it →The order was saved, but the warehouse never saw it
Why can this happen? How would you make the database change and event publication reliable? Why does swapping their order not solve it?
Try it →Integration and APIs (5)
UNKNOWN transaction: who is responsible for closing the case?
Separate ownership of the authoritative outcome, reconciliation, and unresolved operational exceptions.
Try it →New optional API field. Can the application still read the response?
Distinguish an additive schema change from compatibility of running consumers and plan evidence before rollout.
Try it →Should a notification failure block an order?
Separate the business outcome from failure of an auxiliary notification.
Try it →No response. Was the booking created?
Distinguish a missing response from a failed operation.
Try it →The client retried a payment. Will you charge twice?
How will you make retries safe and avoid starting a second charge? What happens when the first call has an unknown outcome?
Try it →Resilience and recovery (4)
Every health check is green. Why can't users complete the journey?
Would you classify the service as available? In 2–3 sentences state its boundary, the signals and controlled test you need, and what would change your decision.
Try it →The result is ready, but saving failed. What do you show?
Preserve the core function when an optional feature fails.
Try it →The database is back; attachments are not. Do you reopen the portal?
What must be restored and checked before you resume submissions? How will you verify a five-minute RPO and 30-minute RTO for the whole service?
Try it →Reports block checkout. What do you isolate?
How will you stop reports from affecting checkout? What limits, overload behaviour and measurements will you define before changing the connection count?
Try it →Analysis and decisions (3)
The provider delivers SaaS. Do you need to know its architecture?
Distinguish vendor implementation details from information needed to assess responsibility, risk and SaaS integration.
Try it →New tablets, the same app. What should you check?
Understand the impact before treating a change as a hardware swap.
Try it →The offer works, but diverges from the architecture requirement
How do you separate requirements compliance from the decision to keep evaluating the offer? What would you recommend, and what evidence is needed before approving the implementation architecture?
Try it →Security (1)
The cache knows the report, but not the tenant. What will tenant B see?
Where did tenant isolation break? How will you change authorisation, the cache key and tests so the fix covers the whole path rather than this endpoint alone?
Try it →