k6 suites that assert the Yemaya creative-studio SLOs from
docs/performance/baselines.md §3.
| Script | Exercises | SLO gate |
|---|---|---|
studio-edit-load.js |
PATCH /v1/projects/:id at interactive cadence |
studio edit p95 ≤ 150 ms |
asset-upload-load.js |
Register → bytes → /assets/:id/complete |
end-to-end upload commit p95 ≤ 2,000 ms |
collaboration-sync-load.js |
POST /v1/collaboration/sessions/join + presence PATCH |
presence update p95 ≤ 250 ms |
Run locally#
bash
docker compose -f docker/docker-compose.dev.yml up -d
# Studio edit round-trip
k6 run testing/performance/k6/yemaya/studio-edit-load.js
# Asset upload (default 10 MB per asset for CI-friendly runs)
k6 run testing/performance/k6/yemaya/asset-upload-load.js
# Multi-user collaboration fan-out
k6 run testing/performance/k6/yemaya/collaboration-sync-load.js
Environment variables#
| Var | Default | Purpose |
|---|---|---|
YEMAYA_API_URL |
http://localhost:3001 |
Yemaya API origin |
JWT_SECRET |
yemaya-load-test-secret |
HS256 secret for load-test tokens |
YEMAYA_EDITS_PER_ITER |
3 |
PATCH ops per VU iteration (keystroke-style batching) |
YEMAYA_ASSET_SIZE_BYTES |
10485760 (10 MB) |
Per-asset upload size. Set to 104857600 for the 100 MB budget. |
YEMAYA_UPLOAD_BUFFER_KB |
256 |
Chunk size when streaming bytes to the signed upload URL |
YEMAYA_HEARTBEATS_PER_ITER |
5 |
Presence heartbeats per collaboration iteration |
Why 10 MB by default instead of 100 MB#
The documented SLO is for uploads up to 100 MB. CI runners can't reliably push
that volume through a shared load-test instance, so the default is 10 MB and the
2 s p95 budget is still the ceiling. For the full 100 MB validation run, set
YEMAYA_ASSET_SIZE_BYTES=104857600 on a dedicated perf runner — the budget does
not change.
What counts as a regression#
Any thresholds block failing in any of the three scripts. k6 exits non-zero
and CI fails. Budget changes require touching both the thresholds in
config.js and the corresponding row in docs/performance/baselines.md §3.