Bot swarm

dev-bypass@local d

Bot swarm

Drive synthetic load against any feature. Pick a scenario, mix personas, watch live latency in real time. No SSH, no CLI, no Prometheus — single pane.

Targeting: http://api:9080

New run

Scenario

Bots

Cap 2000 (server-side hard limit).

Duration (sec)

Don't lead with 1000 bots

Last time a 1000-bot OTP+tournament burst crashed the prior staging box. Ramp gradually so you learn where THIS box saturates:

Smoke 10 bots / 30s first — confirm scenario reaches its endpoints.
Step up to 100 bots / 60s — watch for any 5xx in the table.
500 bots / 60s — measures realistic peak (~3K concurrent users).
1500 bots / 120s — stress the auth burst lane.
Stop if p99 > 2s OR errors > 1% in any row — you've found the ceiling.

Persona mix (% per kind)

realistic

fast

slow

flaky

cheater

Sums > 100 are clamped; remainder spills to realistic.

Fixed seed (optional)

0 = random. Non-zero → deterministic.

API base URL

Notes (optional)

Redis cache experiments

Enable GET /v1/exams cache (5-min TTL, per-locale) Enable GET /v1/leaderboard cache (30-sec TTL, shared envelope per scope/exam — Self stays fresh)

Tags accumulate (cache-on:v1-exams,cache-off:v1-leaderboard), so the compare tab matches the exact flag combination. Pair the leaderboard-browse scenario with the leaderboard checkbox to see the largest delta — that endpoint hits the DB profile-hydration path on every uncached call.

Live

Host load (1m / 5m)

API /v1/health ping

DOWN

Admin goroutines

⚪

Admin mem (heap / sys)

Identity pool authed bots reuse JWTs across runs

cached

access valid

refresh only

Thresholds (2 vCPU): 🟢 healthy · 🟡 strained · 🔴 ceiling found. Load > 2 ⚠ > 4 ✗ · API ping > 300 ms ⚠ > 1 s ✗ · Goroutines > 2k ⚠ > 5k ✗ · p99 > 500 ms ⚠ > 2 s ✗

Past runs (20)

Compare: vs

Click A / B on two rows then hit Compare to see side-by-side diff. Tag your baselines (e.g. "baseline-no-cache") so they're easy to find later.

Run · Tag	Scenario	Bots	Dur	Iter	Err	Phase	Started
run_6158355572c07966	`leaderboard-browse`	1000	60s	44389	30085	done	2026-05-13 19:40
run_7801f7b0a3e580da	`leaderboard-browse`	1000	60s	45251	31661	done	2026-05-13 19:38
run_8e8ae13d1e084bae	`daily-loop`	1000	60s	17046	0	done	2026-05-13 19:35
run_0b58c71ae09ea0cd	`daily-loop`	1000	60s	16427	0	done	2026-05-13 19:33
run_985f661a20d8f985	`daily-loop`	1000	60s	16105	0	done	2026-05-13 19:22
run_74997851f7b9ea42	`pool-warm`	1000	30s	1000	0	done	2026-05-13 19:21
run_8dc4ca45529b8b92	`pool-warm`	1000	120s	1000	0	done	2026-05-13 19:19
run_4c317fecbc969aaf	`auth-flood`	10	60s	168	0	done	2026-05-13 19:09
run_4d4afe13255cd623	`pool-warm`	1000	150s	504	2510	done	2026-05-13 19:02
run_eed6a717b3462a44	`pool-warm`	1000	120s	695	305	done	2026-05-13 17:52
run_4372524269d61646	`pool-warm`	10	60s	10	0	done	2026-05-13 17:50
run_a4787801d283dfa8	`auth-flood`	1000	120s	790	936	done	2026-05-13 17:48
run_56dfe5fc73d5ba44	`auth-flood`	1000	120s	901	906	done	2026-05-13 17:11
run_f41469c718a49b60	`auth-flood`	1000	360s	3473	2042	done	2026-05-13 16:41
run_e337432711279a8a	`auth-flood`	1000	120s	1184	1004	done	2026-05-13 16:38
run_491e57b71a304456	`daily-loop`	250	60s	15358	0	done	2026-05-13 16:33
run_35c461d771471c07	`daily-loop`	250	60s	16254	0	done	2026-05-13 16:31
run_5b6703f3c2a127b3	`daily-loop`	250	60s	11661	0	done	2026-05-13 16:30
run_ffa40eb3606453fb	`daily-loop`	50	60s	5421	0	done	2026-05-13 16:26
run_5c4ceb09a4a6acf2	`daily-loop`	50	60s	5026	0	done	2026-05-13 16:25