Documentation
¶
Overview ¶
Package k6 parses the raw k6 summary that benchmarks/workloads/crud-list.js dumps (via handleSummary) into the load-generator-derived fields of a benchmark result. The workload does no interpretation — it forwards k6's own metrics and state — so every mapping that used to be a footgun in JavaScript (the http_req_failed Rate whose FAILED count is `passes` not `fails`, the percentile keys, the elapsed run time) lives here and is unit-tested against captured k6 goldens (testdata/).
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Summary ¶
type Summary struct {
Requests int64
RequestsPerSecond float64
LatencyMs result.Latency
Errors int64
ChecksFailed int64
DurationSeconds float64
}
Summary is the interpreted, load-generator-measured half of a result row. The orchestrator fills in the identity/config fields (framework, concurrency, trial, ...) around it via Merge. DurationSeconds is the ACTUAL elapsed measured window (from k6's state), not the requested flag.
func ParseSummary ¶
ParseSummary decodes and interprets one raw k6 summary. Decoding/mapping only — validity is Validate's job, so a caller can inspect a failed run's numbers before rejecting it.
func (Summary) Merge ¶
Merge copies the load-generator-measured fields of s onto base (which carries the identity/config fields the orchestrator set), including the actual measured DurationSeconds, and returns the completed result row.
func (Summary) Validate ¶
Validate rejects a trial that isn't a clean measurement: no traffic at all (target unreachable / workload misconfigured), any HTTP error (the target erroring or unreachable — every request failing still yields requests > 0), or any failed content check (a 200 with the wrong page shape). The headline read workload against a healthy, seeded app must be error-free; the orchestrator turns a Validate failure into a loud benchmark failure rather than recording a bogus row (issue #141 §10 "a failed implementation must make the benchmark command fail clearly", §7 "instead of publishing bogus data").
Source Files