RED and USE dashboards

Observability and operations Junior Python 3.13+ PostgreSQL 17+ Reviewed By Best Latest Apps

Pick the fix

Latency p99 jumped but CPU is 30%. Which USE signal should you check next?

Options
Hint 1 of 1

Where can requests wait without using CPU?

Choose an answer, then open Show answer below to check it.

Show answer

Answer: A Saturation of the database connection pool (wait time or waiters)

Correct. Requests may be queueing for a connection.

USE asks about utilisation, saturation and errors for each resource. With CPU low, a queue is the next suspect: requests waiting for a free connection in the pool add latency without using CPU.

Why not the others?
  • B CPU utilisation again. Already low.
  • C Total requests ever served. Not a USE signal.
  • D Disk size. Unlikely to change latency suddenly.

30-second answer

For each request-serving service, RED: Rate (requests per second), Errors (failed requests per second or share), and Duration (latency histogram, shown as p50/p95/p99). For each resource (CPU, memory, disk, connection pools, queues), USE: Utilisation, Saturation (queued work) and Errors. RED tells you users are affected; USE helps find which resource is the cause.

Practise saying it · 25 s. No recording, just a timer

First dashboard.

Say your answer out loud, then check which of these points you covered:

  • RED per service
  • USE per resource
  • Percentiles
  • 4xx vs 5xx
  • Deploy markers

Deep dive

Google's four golden signals (latency, traffic, errors, saturation) cover similar ground.

Saturation often shows first as pool wait time or queue depth.

Split errors into client (4xx) and server (5xx).

Put deploy markers on dashboards.

Common mistakes

  • Dashboards of 60 panels nobody reads.
  • Average latency only.
  • Counting 4xx as server errors.

Senior take & follow-ups

RED for services, USE for resources, and how they combine during an incident.

Follow-up questions

Think of your answer first, then open the hint.

  1. What's saturation?

    Show hint

    Work waiting: queue depth, pool waits, run-queue length.

  2. Why deploy markers?

    Show hint

    Most incidents follow a change.

Related questions

Suggest a fix if something here is wrong or out of date.