Problems / Resilience, hard
The database answers in 70 ms at the median but 350 ms at p99. The API gives up after 150 ms and retries 3 times with no backoff, so the slow tail turns into extra load.
Free to open. Runs in your browser, no sign-up.
Success ≥ 99%, p99 ≤ 1 s.
1,200 requests a second for 30 seconds: 90% reads, 10% writes.
What each component does in this design, and what calls it. The simulator mimics each one from these settings.
| Component | Its job here | Settings |
|---|---|---|
| Load balancerLoad balancer | Traffic enters here.Calls API.Layer 4 balancers route by IP and port; layer 7 balancers read HTTP and can route by path or header. | – |
| APIApp service | Load balancer forwards to it.Calls Postgres.Little's law: throughput ≈ concurrency ÷ latency. 200 slots at 50 ms serve about 4,000 req/s. | 3 replicas × 200 slots · timeout 150 ms |
| PostgresSQL database | API calls it.Postgres runs one process per connection, which is why poolers like PgBouncer sit in front of busy databases. | 4 vCPU · 100 connections |
Try it first. Open one hint at a time if you get stuck.
Compare the API timeout with the database p99.
Count database calls against requests: retries multiply them.
A timeout above the p99 and at most one retry with a budget.
Raise the API timeout above the database p99 ("set the API timeout to 500 ms") and limit retries ("1 retry").