Explain like I'm five · SRE

What is Observability?

Observability, told as a story about looking at the gauge, the log, and the path a request took — all together.

errors upI can see errors went up... but why?the gauge just shows errors went up — it has no idea why

The gauge just shows errors went up — it has no idea why.The number shows up, but the reason doesn't.

error rate 90%this alone doesn't tell me where the problem isa single number alone can't tell you what's actually wrong

A single number alone can't tell you what's actually wrong.You just know the error rate hit 90% — not which ride is the problem.

metrics10:01 OK10:02 OK10:03 ERROR: timeout10:04 ERROR: timeoutlogstracewatch three things together: the numbers, the written log, and the path a request tookthe numbers tell you something's wrong; the log and the path tell you where and why

Watch three things together: the numbers, the written log, and the path a request took.The numbers tell you something's wrong; the log and the path tell you where and why.

Notice it with the numbersthat's metrics
Find it with detailed recordsthat's logs
Trace the path a request tookwhich rides it passed through
Watch all three on one screenthat's a dashboard
metrics10:03 ERROR: timeout10:03 ride=Blogstracenotice it with metrics, find it with logs, confirm which ride it was with traces

Notice it with metrics, find it with logs, confirm which ride it was with traces.The three screens connect by arrow, narrowing down the problem.

STORAGEmetrics, logs, traces — all of itkeep it all forever, cost climbsdelete what's oldkeep only as much as you needkeeping all three forever costs money too — keep only what you need, and delete what's old

Keeping all three forever costs money too.Keep only what you need, and delete what's old.

In one breath

Observability = watching metrics (the numbers), logs (the written record), and traces (the path a request took) together, so you can find not just that something's wrong, but where and why.

The ability to look into a system's internal state through three pillars: metrics, logs, and distributed tracing. Where monitoring answers questions you already knew to ask, observability lets you answer questions you didn't know to ask in advance. A dashboard brings the three pillars together on one screen, and a retention period deletes old data to keep storage costs in check.

When grown-ups say it

Observability
Watching the numbers, the log, and the path together. It lets you answer questions you didn't even know to ask. → the people who keep the park open
Metrics
The bar-graph numbers on the control-room wall. The first thing to tell you something's wrong.
Logs
The detailed log a mechanic keeps. Records exactly when and what happened.
Distributed tracing
The path drawn for one request as it passes through each ride. More important the more rides a request passes through.
Dashboard
The control-room wall where all three show up together. What's invisible apart becomes visible together.
Relation to golden signals
The four gauges are one kind of metric. Observability looks past those gauges, into logs and traces too.
Difference from monitoring
The difference between answering known questions and being able to answer unknown ones. Watching only the gauges is monitoring; digging into logs and traces too is observability.
SLI (service level indicator)
The actual number written on today's operating log. A number picked out from the metrics, used to measure the promise you made.
Next story
How Many Rings Make a Real Emergency →
Want another term as a picture book? Request →