SLIs, SLOs and error budgets
Define an SLI users would recognise, turn an SLO into an error budget, alert on burn rate instead of thresholds, and let the budget decide when to ship.
25 min7 steps
By the end you can
- Choose an SLI that measures user-visible success at the right boundary, with a sensible target
- Compute an error budget in bad minutes from an SLO and its window
- Compute a burn rate and explain why multiwindow burn-rate alerts beat static thresholds
- Apply an error budget policy to release and reliability decisions
Measure what users feel
SLI
good / valid
Latency SLI
99% under 300 ms
proportion, not an average
Measure
near the user
load balancer beats server logs
100% is the wrong target
- 99%432 min
- 99.9%43.2 min
- 99.99%4.3 min
- 99.999%0.43 min
Log scale: each gridline step is 10×.
Turn the SLO into a budget
Window
43,200 min
Budget (0.1%)
43.2 min
30 x 1,440 x 0.001
Loading question…
Burn rate: how fast is the budget going?
- Burn rate 130 days
- Burn rate 65 days
- Burn rate 14.42.1 days
Loading question…
Alert on burn rate, over two windows
Page: fast
- 2% in 1 h (burn 14.4)
- Short window 5 min
Page: medium
- 5% in 6 h (burn 6)
- Short window 30 min
Ticket: slow
- 10% in 3 days (burn 1)
Let the budget decide
Budget left
Ship at your own pace
Large burn
Postmortem
Exhausted
Pause launches; reliability and security only
Window recovers
Resume
Check your mastery
Budget
minutes
window x (1 - SLO)
Burn
error ratio / budget ratio
Score 80% or more. See failure modes and deployments.
Mastery check
4 questions on fresh variants. One attempt each, no time limit. Score 80% or more to master the lesson.
Answer this step to continue.