tests with unspecified duration are not allowed when outputt

Error message

tests with unspecified duration are not allowed when outputting data to k6 cloud

What it means

Raised while creating the cloud output when lib.GetEndOffset reports that the computed execution plan never ends. The k6 Cloud service must know when a test finishes, so with `--out cloud` every scenario needs a bounded duration or end time.

Source

Thrown at internal/cmd/outputs_cloud.go:88

	thresholds := make(map[string][]string)
	for name, t := range test.derivedConfig.Thresholds {
		for _, threshold := range t.Thresholds {
			thresholds[name] = append(thresholds[name], threshold.Source)
		}
	}

	et, err := lib.NewExecutionTuple(
		test.derivedConfig.ExecutionSegment,
		test.derivedConfig.ExecutionSegmentSequence,
	)
	if err != nil {
		return err
	}
	executionPlan := test.derivedConfig.Scenarios.GetFullExecutionRequirements(et)

	duration, testEnds := lib.GetEndOffset(executionPlan)
	if !testEnds {
		return errors.New("tests with unspecified duration are not allowed when outputting data to k6 cloud")
	}

	if conf.MetricPushConcurrency.Int64 < 1 {
		return fmt.Errorf("metrics push concurrency must be a positive number but is %d",
			conf.MetricPushConcurrency.Int64)
	}

	if conf.MaxTimeSeriesInBatch.Int64 < 1 {
		return fmt.Errorf("max allowed number of time series in a single batch must be a positive number but is %d",
			conf.MaxTimeSeriesInBatch.Int64)
	}

	if conf.PushRefID.Valid && conf.PushRefID.String != "" {
		if err := applyExternalProvisioningCreds(gs, test, &conf); err != nil {
			return err
		}
		test.preInitState.RuntimeOptions.Env[testRunIDKey] = conf.PushRefID.String
		gs.Logger.Debug("using PushRefID, skipping CreateTestRun")

View on GitHub (pinned to 93accf6570)

Solutions

Give every scenario an explicit duration, e.g. `{ executor: 'constant-vus', vus: 10, duration: '5m' }`
Or use executors that end on their own budget: arrival-rate with `duration`, shared-iterations/per-vu-iterations with `iterations`
If an open-ended test is intentional, drop `--out cloud` and stream metrics via `--out experimental-prometheus-rw` or another output instead

Example fix

// before
export const options = { vus: 10 };

// after
export const options = {
  scenarios: { smoke: { executor: 'constant-vus', vus: 10, duration: '5m' } },
};

Defensive patterns

Strategy: validation

Validate before calling

// Verify before `k6 cloud` that every scenario is time-bounded
function assertBoundedScenarios(options) {
  const s = options?.scenarios ?? { default: options };
  for (const [name, sc] of Object.entries(s)) {
    const bounded = sc.duration || sc.startTime || ['shared-iterations', 'per-vu-iterations'].includes(sc.executor);
    if (!bounded) throw new Error(`scenario '${name}' has unspecified duration; cloud output requires it`);
  }
}
assertBoundedScenarios(options);

Prevention

Always give scenarios an explicit `duration` (or endTime) when cloud output may be used
Smoke-test `k6 inspect script.js` locally before `k6 cloud` to see the computed execution plan
Keep a CI lint that rejects options files where an executor lacks duration/iterations

When it happens

Trigger: `k6 run --out cloud` (or `k6 cloud`) with an executor whose duration is unspecified, e.g. `constant-vus`/`ramping-vus` stages without `duration`, or an options set whose overall plan leaves `testEnds` false.

Common situations: Reusing locally-tuned options (open-ended scenarios, only `vus` set) with the cloud output; refactors that removed `duration` from a scenario; CLI overrides like `--vus 10` without `--duration` while cloud output is enabled.

Related errors

AI-assisted analysis of grafana/k6@93accf6570 (2026-08-15). Data as JSON: /api/errors/eb318e69a4a3bf91. Report an issue: GitHub.