DistrLab v1.1.0 — circuit breaker & smarter load balancing

Resilience depth

Everything from v1.0's seven scenarios, plus deeper failure handling.

Circuit breaker

• New Circuit breaker scenario (#8): users → breaker → flaky catalog → postgres.
• Breaker trips OPEN past an error-rate threshold, fast-fails with 503, goes half-open after a cooldown, and closes once its probes succeed.
• Link down toggle + flakiness + trips-at / cooldown / probe-limit knobs.
• Inspector Trip / Reset buttons; OPEN / HALF-OPEN badges on the topology.
• Measured: link down → p95 collapses from ~4020 ms to ~20 ms; repair → "Healthy" returns automatically.

Smarter load balancing

• Power of two choices (P2C) and least-latency (response-time EMA) join round-robin, random, least-inflight and sticky.

Fixes & quality

• Health flags now recover after an outage (intermediate nodes are no longer stuck "unhealthy" forever).
• New headless engine test suite: bun test → 20 passing.
• New ROADMAP.md (v1.0 → v2.0); README updated for eight scenarios.

Live at https://distrlab.onrender.com

DistrLab v1.0.0

DistrLab — a distributed-systems playground that runs entirely in your browser.

Break a live architecture and watch it recover: cache stampedes, queue backpressure, rate limiting, load-balancer strategies, network partitions, replication lag. Tune knobs, inject faults (crash / slow / expire / partition), and read plain-language explanations of what is happening.

Free-tier by design — no backend, no LLM, no paid APIs; a Web Worker on a static page, deployable to any static host.

Scenarios

#	Scenario	Lesson
1	Playground	healthy baseline: p95, throughput, capacity, queues, cache hit ratio
2	Cache stampede	single-flight off ⇒ every expiry is a thundering herd on the catalog service
3	Queue backpressure	shallow queues shed fast; deep queues trade errors for latency
4	Rate limiting	a token bucket protects the service at the cost of 429s
5	Load balancer strategies	round-robin / random / least-inflight / sticky under skew and failure
6	Network partition	health checks + least-inflight route around a cut link
7	Replication lag	stale replica reads versus primary reads

Controls: 1–7 switch scenarios · space pause · R reset · click a node for the inspector (crash/slow/expire/partition) · knobs tune arrival/capacity/TTL/single-flight/LB/health/timeouts/lag.

Verifications

• 12/12 headless engine tests over all scenarios
• svelte-check clean; production build ~79 kB JS + 29 kB worker (~33 kB gzipped)
• headless-browser smoke test: mount, canvas render, scenario switch, inspector, transport controls — zero runtime errors

Demo GIF, screenshots, and render.yaml (static-site deployment) attached.