🔒 Paid

Network Partition

Inter-region replication between Region A and Region B has been severed, while each region keeps serving traffic normally with full local capacity. The business draws a hard line between two kinds of data: account-ledger writes (balances, financial transactions) must never be accepted by more than one region at the same time — a divergent ledger write during the partition is a critical, unrecoverable correctness incident no matter how healthy write availability looks in either region. Profile data (display name, UI preferences, notification settings) carries no such requirement and may legitimately tolerate temporary staleness, or reduced availability in one region, for the duration of the partition. Decide how ledger writes should be routed while the regions can't talk to each other, and separately decide how much availability you are willing to trade for freshness on profile data.

Difficulty
Staff
Expected Time
~20 min
Category
Distributed Systems

Skills

  • Recognizing when customer-facing availability conceals a correctness failure
  • Encoding different consistency requirements for different data categories
  • Distinguishing a structural write-routing fix from a manually-achieved safe state
  • Reasoning about a genuine, bounded consistency-for-availability tradeoff

System Architecture

  • Region A
  • Replicated Service
  • Region B

This is a Pro scenario. Upgrade to Pro to unlock it.

Upgrade to ProBack to Incidents