AO3 Mirror v4 — initial commit

Cookie-aware proxy pool with CF challenge solving
- Tiered proxy pool (fast 50 + main 676)
- Per-proxy cf_clearance cookie persistence
- CF challenge detection + user-browser solving
- Safari + Chrome TLS fingerprint rotation
- Async FastAPI backend with LRU cache
- Passive daemon with systemd supervision
- Stats dashboard + Prometheus metrics
This commit is contained in:
akiba
2026-06-23 08:04:04 +00:00
commit 085cc7a140
16 changed files with 2842 additions and 0 deletions

146
scripts/daemon.py Normal file
View File

@@ -0,0 +1,146 @@
#!/usr/bin/env python3
"""
AO3 Mirror - 被动守护进程 (Daemon v2)
不再主动 e2e 检查或重启 healthy worker。
职责:
1. 每 30s 检查 worker 是否活着,只有连续 3 轮都挂掉才重启
2. 不做 e2e不做 caddy 检查(让 systemd 管)
3. 每 5min 写状态文件
"""
import json
import os
import subprocess
import sys
import time
import logging
import signal
BASE_DIR = "/home/ubuntu/ao3-mirror"
SHM_DIR = "/dev/shm/ao3"
WORKER_PORTS = [8081, 8082]
CHECK_INTERVAL = 30
STATUS_FILE = f"{SHM_DIR}/status.json"
os.makedirs(SHM_DIR, exist_ok=True)
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s [DAEMON] %(message)s",
handlers=[logging.StreamHandler()],
)
logger = logging.getLogger("ao3-daemon")
state = {
"started_at": time.time(),
"consecutive_dead": {p: 0 for p in WORKER_PORTS},
"restarts_today": 0,
"alerts": [],
}
def _http_get(url: str, timeout: int = 5) -> tuple[bool, str]:
import urllib.request
try:
req = urllib.request.Request(url)
resp = urllib.request.urlopen(req, timeout=timeout)
return (resp.status == 200, resp.read().decode("utf-8", errors="replace")[:200])
except Exception as e:
return (False, str(e)[:200])
def check_worker(port: int) -> bool:
ok, body = _http_get(f"http://127.0.0.1:{port}/health", timeout=4)
if ok:
try:
return json.loads(body).get("status") == "ok"
except Exception:
pass
return False
def restart_worker(port: int):
logger.warning(f"Restarting worker on port {port}")
state["restarts_today"] += 1
subprocess.run(["fuser", "-k", f"{port}/tcp"], capture_output=True, timeout=10)
time.sleep(1)
python_bin = "/home/ubuntu/.hermes/hermes-agent/venv/bin/python3"
i = WORKER_PORTS.index(port)
log_file = f"{BASE_DIR}/worker-{i}.log"
proc = subprocess.Popen(
["taskset", "-c", str(i), python_bin, "-m", "uvicorn", "app:app",
"--host", "127.0.0.1", "--port", str(port), "--workers", "1",
"--loop", "uvloop", "--http", "httptools",
"--log-level", "warning", "--timeout-keep-alive", "30"],
cwd=BASE_DIR, stdout=open(log_file, "a"), stderr=open(log_file, "a"),
)
pid_file = f"{SHM_DIR}/worker-{i}.pid"
with open(pid_file, "w") as f:
f.write(str(proc.pid))
logger.info(f"Worker {port} restarted (PID: {proc.pid})")
def check_workers():
"""Only restart if worker has been dead for 3 consecutive checks (90s)."""
for port in WORKER_PORTS:
if check_worker(port):
state["consecutive_dead"][port] = 0
else:
state["consecutive_dead"][port] += 1
if state["consecutive_dead"][port] >= 3:
restart_worker(port)
state["consecutive_dead"][port] = 0
dead_count = sum(1 for p in WORKER_PORTS if state["consecutive_dead"][p] > 0)
alive = len(WORKER_PORTS) - dead_count
if dead_count > 0:
logger.info(f"{alive}/{len(WORKER_PORTS)} workers alive ({dead_count} degraded)")
def save_status():
worker_status = {}
for i, port in enumerate(WORKER_PORTS):
worker_status[f"worker_{i}"] = {
"port": port, "healthy": check_worker(port), "uptime_s": int(time.time() - state["started_at"]),
}
data = {
"timestamp": time.time(), "uptime": int(time.time() - state["started_at"]),
"workers": worker_status, "restarts_today": state["restarts_today"],
}
with open(STATUS_FILE + ".tmp", "w") as f:
json.dump(data, f)
os.rename(STATUS_FILE + ".tmp", STATUS_FILE)
def handle_signal(signum, frame):
logger.info(f"Signal {signum}, exiting")
sys.exit(0)
def main():
signal.signal(signal.SIGTERM, handle_signal)
signal.signal(signal.SIGINT, handle_signal)
logger.info("Daemon v2 started (passive mode)")
logger.info(f"Watching {WORKER_PORTS}, check every {CHECK_INTERVAL}s, restart after 3 missed checks")
# Initial startup: start all workers immediately
for port in WORKER_PORTS:
if not check_worker(port):
restart_worker(port)
time.sleep(2)
state["consecutive_dead"][port] = 0
time.sleep(5) # Let them settle
last_status = 0
while True:
try:
check_workers()
now = time.time()
if now - last_status >= 300:
save_status()
last_status = now
except Exception as e:
logger.error(f"Loop error: {e}")
time.sleep(CHECK_INTERVAL)
if __name__ == "__main__":
main()

View File

@@ -0,0 +1,25 @@
#!/usr/bin/env python3
"""Restart both AO3 workers manually."""
import subprocess, time, os
BASE_DIR = "/home/ubuntu/ao3-mirror"
SHM_DIR = "/dev/shm/ao3"
VENV = "/home/ubuntu/.hermes/hermes-agent/venv/bin/python3"
os.makedirs(SHM_DIR, exist_ok=True)
for i, port in [(0, 8081), (1, 8082)]:
log = f"{BASE_DIR}/worker-{i}.log"
proc = subprocess.Popen(
["taskset", "-c", str(i), VENV, "-m", "uvicorn", "app:app",
"--host", "127.0.0.1", "--port", str(port), "--workers", "1",
"--loop", "uvloop", "--http", "httptools",
"--log-level", "warning", "--timeout-keep-alive", "30"],
cwd=BASE_DIR, stdout=open(log, "a"), stderr=open(log, "a"),
)
pid_file = f"{SHM_DIR}/worker-{i}.pid"
with open(pid_file, "w") as f:
f.write(str(proc.pid))
print(f"Worker {i} port {port} started PID={proc.pid}")
time.sleep(2)
print("Both workers started")

View File

@@ -0,0 +1,131 @@
#!/usr/bin/env python3
"""
Optimized proxy scanner — uses port-filtered Webshare proxies for 72.6% success rate.
Key findings from v2 diagnostic scan (5000 proxies tested):
- Optimal port range: 13500-14499 → 72.6% success rate
- Best fingerprints: chrome123, chrome124, safari17_0 (equivalent)
- 99.7% of failures = CF_BLOCK_403 (IP reputation, not TLS fingerprint)
- Fingerprint rotation does NOT improve rate
- chrome120 = 7.5%, edge120 = 0% (DO NOT USE)
"""
import concurrent.futures
import os
import sys
import time
from curl_cffi import requests as curl_requests
TEST_URL = "https://archiveofourown.org/robots.txt"
TIMEOUT = 12
MAX_WORKERS = 50
PROXY_FILE = "/home/ubuntu/proxy.txt"
OUTPUT_FILE = "/dev/shm/working_proxies.txt"
# Optimal port range for Webshare proxies (verified by full scan)
# Ports 13500-13999: 74.4% | Ports 14000-14499: 70.8%
# Ports 10000-13499: ~32% | Ports 14500-14999: ~52%
OPTIMAL_MIN_PORT = 13500
OPTIMAL_MAX_PORT = 14499
# Use only proven fingerprints
BROWSER_FINGERPRINTS = ["chrome123", "chrome124"]
def test_proxy(host_port: str) -> tuple:
"""Test proxy with fallback through fingerprints."""
start = time.time()
for imp in BROWSER_FINGERPRINTS:
try:
session = curl_requests.Session()
session.impersonate = imp
session.timeout = TIMEOUT
session.proxies = {
"http": f"http://{host_port}",
"https": f"http://{host_port}",
}
session.headers.update({
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/121.0.0.0 Safari/537.36",
"Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8",
"Accept-Language": "en-US,en;q=0.9,zh-CN;q=0.8",
})
resp = session.get(TEST_URL)
elapsed = time.time() - start
session.close()
if resp.status_code == 200:
return (host_port, True, round(elapsed, 2), f"OK({imp})")
elif resp.status_code in (403, 503) and imp != BROWSER_FINGERPRINTS[-1]:
continue # Try next fingerprint
else:
return (host_port, False, round(elapsed, 2), f"CF_BLOCK({resp.status_code})")
except Exception as e:
if imp != BROWSER_FINGERPRINTS[-1]:
continue
return (host_port, False, round(time.time() - start, 2), str(e)[:60])
return (host_port, False, round(time.time() - start, 2), "ALL_FAILED")
def main():
print(f"[Port-Optimized Scanner] Loading proxies from {PROXY_FILE}")
proxies = []
with open(PROXY_FILE) as f:
for line in f:
line = line.strip()
if not line or ":" not in line:
continue
if "|" in line:
line = line.split("|")[-1].strip()
proxies.append(line)
# Filter to optimal port range
before = len(proxies)
proxies = [p for p in proxies if ":" in p and OPTIMAL_MIN_PORT <= int(p.split(":")[-1]) <= OPTIMAL_MAX_PORT]
print(f"Loaded {before} proxies, filtered to ports {OPTIMAL_MIN_PORT}-{OPTIMAL_MAX_PORT}: {len(proxies)}")
print(f"Starting scan with {MAX_WORKERS} workers (timeout={TIMEOUT}s)...")
working = []
checked = 0
start_time = time.time()
with concurrent.futures.ThreadPoolExecutor(max_workers=MAX_WORKERS) as executor:
futures = {executor.submit(test_proxy, p): p for p in proxies}
for future in concurrent.futures.as_completed(futures):
host_port, is_ok, elapsed, note = future.result()
checked += 1
if is_ok:
working.append((host_port, elapsed))
working.sort(key=lambda x: x[1])
if checked % 100 == 0 or is_ok:
elapsed_total = time.time() - start_time
rate = checked / elapsed_total if elapsed_total > 0 else 0
eta = (len(proxies) - checked) / rate if rate > 0 else 0
status = "" if is_ok else ""
print(f" [{checked}/{len(proxies)}] {status} {host_port} ({elapsed:.1f}s) {note} | "
f"Working: {len(working)} ({len(working)/checked*100:.1f}%) | "
f"{elapsed_total:.0f}s | ETA: {eta:.0f}s")
with open(OUTPUT_FILE, "w") as f:
for host_port, elapsed in working:
f.write(f"{host_port}\n")
elapsed_total = time.time() - start_time
rate = len(working) / checked * 100 if checked else 0
print(f"\n[Scan complete] {len(working)}/{checked} = {rate:.1f}% working (target: 70%+)")
print(f" Baseline (unfiltered): 44.8%")
print(f" Improvement: +{rate - 44.8:.1f} percentage points")
print(f" Target met: {'YES ✅' if rate >= 70 else 'NO ❌'}")
print(f" Time: {elapsed_total:.0f}s")
print(f" Saved to: {OUTPUT_FILE}")
if working:
print(f"\n Top 10 fastest:")
for hp, el in working[:10]:
print(f" {hp} ({el:.2f}s)")
if __name__ == "__main__":
main()