What is the mechanism for validating data integrity after a load test run in k6

k6 does not validate data integrity against an external data source after a load test run completes. Its assertion mechanism operates during execution: check() evaluates response properties and records pass/fail metrics, but it does not query databases or external services once iterations finish. For post-test integrity analysis, the standard mechanism is the built-in JSON output captured with --out json=results.json, paired with separate external tooling to query the target data store.

Cross-request state validation. Within a single virtual user (VU), you can validate state across requests by storing response values in local variables and passing them into subsequent check() calls. Cross-VU state is different: VUs are isolated by default, so shared state requires a SharedArray or an external module. Neither approach lets assertions query an external data source directly after the run.

Output format. JSON is the standard k6 output for post-test integrity checks because it preserves per-check metrics, request timings, and tags in a structured file that external scripts can parse. Other outputs exist, but JSON is the most straightforward for offline integrity comparison.

Steps for this case

1. Use check() during the run to validate response payloads and capture identifiers (IDs, tokens, timestamps) needed for later comparison.
2. Export the run with k6 run --out json=results.json script.js.
3. Parse the JSON to extract expected values, then run separate database or API queries to confirm the data source reflects the expected state.

Verification

Run a minimal script with a check() and the JSON output flag, then inspect the file for check metrics:

k6 run --out json=results.json script.js

Confirm the JSON contains entries for each check. For external validation, use a database client to query the target store and compare counts or records against the identifiers captured during the run.

Thresholds can fail the overall test when check failure rates exceed a limit, but they only evaluate metrics collected during the run; they do not validate persisted data.

This assumes a standard k6 distribution where JSON output is built in. If your requirement is strictly post-run validation against an external database, k6 alone is insufficient—plan for an external verification script. If you only need to confirm that responses were correct during the load test, check() with JSON output is the correct mechanism.