Report format¶

Every run writes report.json and, beside it, report.html (-out and -html choose the paths). -junit PATH also writes JUnit XML. Reading a report explains how to use them; this page lists the fields.

`report.json`¶

{
  "schemaVersion": 2,
  "name": "cashier-charge-must-survive",
  "startedAt": "2026-10-29T13:58:34.98+05:30",
  "elapsed_ns": 15200000000,
  "passed": false,
  "recovery_ns": 199250833,
  "phases": [ ... ],
  "hypotheses": [ ... ],
  "faultLog": [ ... ],
  "probes": [ ... ],
  "timeseries": [ ... ],
  "observedVsInjected": [ ... ],
  "warnings": [ ... ]
}

Field	Meaning
`schemaVersion`	the report format version, currently `2`
`name`	`metadata.name` of the scenario
`startedAt`, `elapsed_ns`	when the run started, and how long it took (nanoseconds)
`passed`	`true` when every hypothesis held
`recovery_ns`	time from the end of the faults to healthy again
`phases`	one summary per phase: `steady`, `inject`, `recover`
`hypotheses`	each hypothesis with its verdict
`faultLog`	each fault: `name`, `kind`, `start`, `end`, `intensity`, `rollback` (`ok` or `fail: ...`) and its parameters as written (`tunables`)
`probes`	every probe sample
`timeseries`	per-second buckets for the charts
`observedVsInjected`	whether each fault visibly landed
`warnings`	everything the run warned about
`load`, `loadGenerator`	the load generator's output, when `load` is set
`prometheus`	the Prometheus series, when `observability.prometheus` is set

Durations whose key ends in _ns are nanoseconds.

A phase¶

{
  "name": "inject",
  "from": "2026-10-29T13:58:37.98+05:30",
  "to": "2026-10-29T13:58:45.98+05:30",
  "errorRate": 0.9756,
  "availability": 0.0244,
  "p50_ns": 279000,
  "p95_ns": 353875,
  "p99_ns": 421834,
  "rps": 5.12,
  "samples": 41,
  "failed": 40
}

errorRate and availability are fractions (0 to 1) here; hypotheses state them as percentages.

A hypothesis¶

{
  "name": "charge survives its own outage",
  "metric": "error_rate",
  "operator": "<",
  "limit": "5%",
  "observed": "97.56%",
  "passed": false
}

observed is "no data" with "noData": true when the window had no samples, and "not recovered" for a recovery hypothesis whose service never came back. allowNoData is copied from the scenario.

Versions¶

A report without schemaVersion was written by an older litmus-lite and is read as version 1. Later versions only add fields. litmus-lite refuses to read a report newer than it knows, and says so.

Version	Change
1	no `schemaVersion`
2	adds `schemaVersion`, and `noData` and `allowNoData` on each hypothesis

`report.html`¶

One self-contained file with no external requests: open it from disk or attach it to a CI run. Sections: the verdict, fault against impact (a chart of the injected intensity, p95 and error rate over time, with the inject window shaded), hypotheses, a phase comparison table, observed against injected, the fault log and a probe summary.

JUnit XML¶

One testcase per hypothesis (class litmus-lite.hypothesis), with a failure that carries the metric, operator, limit and observed value. A scenario with no hypotheses is one test case for the whole run.

<testsuite name="cashier-charge-must-survive" tests="2" failures="1">
  <testcase name="charge survives its own outage" classname="litmus-lite.hypothesis">
    <failure message="error_rate &lt; 5%">observed 97.56%</failure>
  </testcase>
  <testcase name="recovers" classname="litmus-lite.hypothesis"></testcase>
</testsuite>