elemes/load-test

Elemes Load Test

Load test E2E menggunakan Locust yang otomatis di-generate dari folder content/.

Cara Pakai

cd elemes/load-test

# 0. Virtual Environment
python3 -m venv ./env
source ./env/bin/activate

# 1. Install dependency
pip install -r requirements.txt

# 2. Generate test data dari content/ (konten ada di subfolder dasar/, arduino/, circuit/)
#    Token sintetis LOCUST_TEST_* otomatis di-seed ke PostgreSQL bila env DATABASE_URL ter-set
export DATABASE_URL=postgresql+psycopg://elemes:<password>@127.0.0.1:5432/elemes
python content_parser.py --content-dir ../../content --num-tokens 50

# 3. Jalankan Locust (opsional: set VELXIO_HOST jika Velxio bukan di localhost:8001)
export VELXIO_HOST=http://localhost:8001
locust -f locustfile.py

Buka http://localhost:8089, masukkan URL backend Elemes (misalnya http://localhost:3000), lalu mulai test.

File

File	Fungsi
content_parser.py	Parse content/**/*.md (rekursif subfolder) → test_data.json + seed token LOCUST_TEST_* ke PostgreSQL
locustfile.py	Locust script (task weighted) yang baca test_data.json
test_data.json	Auto-generated, jangan di-commit
requirements.txt	Dependency (locust)
e2e_session_check.py	Smoke-test alur interaktif (PTY session) via Flask proxy — jalan di container

Test Scenarios (9 Tasks)

#	Task	Weight	Target	Deskripsi
1	Browse Lessons	3	Elemes	GET /lessons, validasi jumlah lesson
2	View Detail	5	Elemes	GET /lesson/{slug}.json, validasi field per tipe
3	Compile C	4	Elemes	POST /compile, validasi output vs expected
4	Compile Python	3	Elemes	POST /compile, validasi output vs expected
5	Interactive Session Python	4	Elemes	POST /compile/sessions → poll prompt → POST /input → verifikasi output → DELETE
6	Verify Arduino	2	Elemes	GET lesson Arduino, validasi JSON structure
7	Complete Flow	2	Both	fetch → compile → track-progress
8	Progress Report	1	Elemes	Login guru → GET /progress-report.json
9	Compile Arduino	3	Velxio	POST /api/compile, validasi hex_content

Task 5 (interactive session) menguji alur baru playground: Run → prompt muncul → ketik jawaban → Enter → output. Skrip memakai token (login) agar lolos rate-limit anonymous dan benar-benar mengukur kapasitas compiler worker (max 50 sesi, 2 compile bersamaan, queue timeout 20 s).

Load Test 50 Pengguna (2 Device)

Host hanya 4-core/3.5 GiB — 50 kompilasi berat bersamaan tidak realistis, tapi 50 sesi yang menunggu input aman (proses idle tidak makan CPU). Batas worker: INTERACTIVE_MAX_SESSIONS=50, INTERACTIVE_MAX_COMPILES=2, INTERACTIVE_QUEUE_TIMEOUT_SECONDS=20.

Device A — Locust Master (jalankan UI di sini)

cd elemes/load-test
source ./env/bin/activate
locust -f locustfile.py --master --web-port 8089

Device B (atau beberapa device) — Locust Worker

cd elemes/load-test
source ./env/bin/activate
# Ganti <IP_MASTER> dengan IP device A (di jaringan/Tailscale yang sama)
locust -f locustfile.py --worker --master-host <IP_MASTER>

Mulai Test

1. Buka http://<IP_MASTER>:8089 di browser device A.
2. Host: URL app yang diuji, misal https://sinau-c-dev.manakin-gentoo.ts.net (Tailscale Funnel) atau http://localhost:3000.
3. Number of users: 50 · Spawn rate: 5–10/s (ramp-up ~5–10 detik agar hampir bersamaan) · Duration: 5–10 menit.
4. Amati di tab Charts: RPS per endpoint, response time /compile/sessions & /input, dan failure rate.
5. Skenario /compile/sessions [Python interactive] wajib ≥95% sukses; failure 429 pada create menandakan anon rate limit (bukan kapasitas) — gunakan token.

Catatan: task compile biasa (/compile) tetap kena anon rate limit 1 per 2 menit bila token tidak ikut — login di on_start menyimpan cookie, jadi user ber-token exempt.

Re-generate Setelah Tambah Lesson Baru

Setiap kali ada lesson baru di content/, cukup jalankan ulang:

python content_parser.py --content-dir ../../content --num-tokens 50

test_data.json akan di-update otomatis dan Locust langsung test lesson baru.

E2E Smoke Test (container)

Verifikasi cepat alur interaktif tanpa browser — jalan di dalam container backend:

# Nama container mengikuti folder deploy (PROJECT_NAME = basename folder
# parent, mis. lms-dev → lms-dev_elemes_1) — resolve via label project+service,
# jangan hard-code nama container:
ELEMES_CTN=$(podman ps -q \
  --filter "label=io.podman.compose.project=<PROJECT_NAME>" \
  --filter "label=com.docker.compose.service=elemes")
podman cp elemes/load-test/e2e_session_check.py "$ELEMES_CTN":/tmp/
podman exec -e E2E_TOKEN=<token_siswa> "$ELEMES_CTN" python3 /tmp/e2e_session_check.py
# Contoh token: LOCUST_TEST_xxxx — token sintetis yang di-seed ke PostgreSQL
# oleh content_parser.py (jalankan langkah 2 dengan DATABASE_URL ter-set).
# <PROJECT_NAME> = nama folder deploy (mis. lms-dev / sinau-c).

Tanpa E2E_TOKEN skrip tetap jalan tapi hanya 2 create session (anon rate limit 1/2 menit).