# Failure and recovery evidence

Compare fault detection, routing events, and observed service behavior on the same clock. A route change is an event. A working application is an observation.

## Try these decisions

1. In the first scenario, compare the route change 0.7 seconds after the declared fault with the first successful probe 6 seconds after it.
2. Choose the two-series scenario. Ping can pass while the application still fails. The first successful application transaction arrives after 10 seconds in this invented example.
3. Choose a gap with no observed failure. Explain why missing observations cannot establish a measured outage duration.
4. Choose repeated failures. Decide whether the first passing observation supports closing the incident.
5. In the controlled sampling exercise, application failure is declared at 10 seconds and recovery at 15.4 seconds. Change the sample interval and offset while keeping that truth fixed. Predict how the observed recovery bracket moves.

## Interpret an interval honestly

A failing sample followed by a passing sample brackets an observed failure-to-success transition between those samples. The exact transition time is unknown. It does not establish uninterrupted health between samples or after the passing observation. Analyze each series separately: a passing probe cannot certify an application transaction.

Ordinary uploaded observations do not include the hidden truth from the controlled experiment. Detection and route-change timestamps alone do not prove service recovery. A declared fault is not proof of the cause.

## Work with observations

Use synthetic or non-sensitive, time-ordered CSV data. The browser lab accepts at most 2,000 samples and 12 series. Analyze the observations, inspect the report, and save a copy. Nothing is uploaded and no router is contacted.

## Limits

The lab interprets supplied samples; it is not a network simulator or a live monitoring system. Gaps do not establish outages, separate series do not certify each other, and later failures still matter after an earlier success.

## Continue

- [Open the interactive lab](https://oaktowntech.com/wp-content/uploads/oaktown-field-notes-v76/failure-evidence.html)
- [Browse all labs](https://oaktowntech.com/labs/)
- [Read Field Notes](https://oaktowntech.com/field-notes/)
