Safari Reading List Wiki Home About Schema Log Sources

Concepts

Distributed Systems Correctness medium

Type
concept
Tags
data-systems software-engineering
Confidence
medium
Created
2026-08-09
Updated
2026-08-09
Sources
raw/articles/why-apache-kafka-doesn-t-need-fsync-to-be-safe-jack-vanlightly-2cb7e0565013.md, raw/articles/how-to-do-distributed-locking-martin-kleppmann-s-blog-b7e13681b8fa.md, raw/articles/how-to-build-a-distributed-queue-in-a-single-json-file-on-object-storage-ada83d75ce43.md, raw/articles/we-replaced-redis-with-mysql-for-inventory-reservations-and-it-scaled-20-e331d101bf80.md, raw/articles/how-when-aws-was-down-we-were-not-authress-knowledge-base-40afcfe9e0a0.md

Distributed Systems Correctness

Across queues, logs, locks, reservations, and failover systems, the corpus distinguishes a fast common path from the conditions that make a state change safe under failures. Correctness comes from explicit protocol guarantees, bounded scopes, and recoverable records—not from assuming a cache, lease, or network stays healthy.

Synthesis

  • Kafka’s recovery protocol and Kleppmann’s fencing-token argument both show why durability, leadership, and stale writers must be analyzed together rather than by a single write primitive. [src] [src]
  • The object-storage queue and Shopify reservation designs use conditional mutation, acknowledged durable state, heartbeats or locks, and cautious rollout to preserve useful semantics in the presence of retries and concurrency. [src] [src]
  • Authress’s regional failover account expands the boundary outward: health checks should test dependencies and business logic, while a five-nines target remains a design goal rather than a guarantee. [src]

Related pages