"""压并发：12 条同时投给调度层，看三件事。

1. 派发会不会打偏——调度层是「读 health 挑最闲的」，同时提交时几个请求可能读到同一份快照。
   顺序提交测不出这个，必须真并发。
2. queue_position 和 eta_seconds 在排队时说了什么。
3. 队列有没有上限、超了会怎样。
"""
import json, time, urllib.request
from collections import Counter
from concurrent.futures import ThreadPoolExecutor

GW = "http://127.0.0.1:8000"
N = 12
SPEC = {"prompt": "一只橘猫趴在洒满阳光的窗台上，慢慢眨眼", "frames": 22, "steps": 4, "turbo": True}


def call(path, data=None, timeout=60):
    req = urllib.request.Request(GW + path, data=data,
                                 headers={"Content-Type": "application/json"} if data else {})
    with urllib.request.urlopen(req, timeout=timeout) as r:
        return json.load(r)


def submit(i):
    t = time.time()
    job = call("/v1/videos", json.dumps(dict(SPEC, seed=100 + i)).encode())
    return i, job, time.time() - t


t0 = time.time()
with ThreadPoolExecutor(max_workers=N) as pool:
    results = sorted(pool.map(submit, range(N)))
print(f"12 条并发提交耗时 {time.time() - t0:.2f}s\n")

for i, job, dt in results:
    print(f"  #{i:<2} {job['backend'][-4:]}  queue_position={job['queue_position']}  "
          f"eta={job['eta_seconds']}s  ({dt:.2f}s)")

spread = Counter(job["backend"][-4:] for _, job, _ in results)
print(f"\n派发分布: {dict(sorted(spread.items()))}  理想是每个 {N // 4} 条")

ids = [job["id"] for _, job, _ in results]
print("\n--- 轮询")
while True:
    time.sleep(30)
    states = [call(f"/v1/videos/{jid}") for jid in ids]
    c = Counter(s["status"] for s in states)
    print(f"  {dict(c)}", flush=True)
    if all(s["status"] in ("done", "failed") for s in states):
        break

total = time.time() - t0
print(f"\n12 条全部完成，墙钟 {total:.0f}s")
for s in sorted(states, key=lambda s: s["created_at"]):
    took = s["finished_at"] - s["started_at"]
    wait = s["started_at"] - s["created_at"]
    print(f"  {s['id'][:6]} {s['status']}  排队{wait:5.0f}s  执行{took:4.0f}s  {s['backend'][-4:]}")