CoursesSystem Design Foundations
Reliability and Consistency
Timeouts and the retry budget, failover and the writes it can lose, the operations that need strong consistency and the ones that do not, and the duplicate delivery every queue eventually makes.
- Lessons
- 6
- Exercises
- 35
- Minutes
- 38
- 1
Timeouts and Retry Budgets
After this lesson you can set a timeout from the target it protects, count what naive retries do to a slow dependency, and write a retry that helps instead of piling on.
DebugMultiple choiceShort answerTrace - 2
Failover Without Losing Writes
After this lesson you can say what a failover loses under asynchronous replication, when to pay for synchronous replication instead, and what the health check must decide before promoting a replica.
Fill blankMultiple choiceShort answerTrace - 3
Strong Where It Counts
After this lesson you can say which operations must read the latest write and which may read a slightly old one, and pay for strong consistency only where a stale read is a wrong decision.
Fill blankMultiple choiceTapTrace - 4
Duplicate Delivery Is Normal
After this lesson you can build a consumer that survives at-least-once delivery, keep order where order matters, and route the message that can never succeed somewhere an operator will see it.
DebugMultiple choiceTrace - 5
Checkpoint: Reliability and ConsistencyCheckpoint
Timeouts, failover, consistency and duplicates for a system you have not seen before.
DebugMultiple choiceShort answerTrace - 6
Boss: Inventory ReservationBoss
One design carried from the ask to the post-mortem: the question, the estimate, the reservation contract, where the count lives, the retry storm, the failover that oversold, the revised write path, and the tradeoff you have to defend.
Code orderMultiple choiceShort answerTrace