What is an SLO?
A Service Level Objective, told as a story about the line a park draws for itself and promises to stay inside.
The control room and the mechanic argue over this every time.Everyone's idea of fine is different, so it always ends in a fight.
Why is this hard?Without a shared line, everyone's idea of fine is different.
Look at today's numbers, then promise a target in advance.Stay inside it freely — cross it, and everyone pauses.
The real numbers wander above and below the target line.When a stretch crosses the line, you can see it right away.
There's no such thing as perfect uptime. Instead, promise in advance.Set the target too high, and the cost never stops growing.
An SLO is the target line a team sets for itself by looking at today's record (SLI). Stay inside it and move freely; cross it, and fixing comes before any new feature.
A target the team sets for itself, distinct from the formal SLA made with customers — usually stricter than the SLA. The slack left over is the error budget, and a burn rate tracks how fast it's being spent.
When grown-ups say it
- SLO
- The promise line a team sets for itself. Cross it, and everyone pauses.
- SLI
- The number actually measured today. This number is what sets the target. → today's record board
- Error budget
- The slack left over by the target. Spend it on bold new tries. → the bundle of allowed-failure tickets
- Setting the target
- Starts from today's record. Set it too high, and the cost never stops growing.
- SLO vs SLA
- What's set internally versus promised to the customer. The contract is usually looser than this. → the promise written up for the customer
- Review cadence
- Looked at again every so often. Because both guests and the system change.
- Burn rate
- How fast the slack gets spent. A fast drop triggers a warning early.
- Golden signals
- The four gauges that measure today's record. Latency, traffic, errors, saturation. → the four gauges on the control-room wall