claudetools/clients/internal-infrastructure/scripts/cloudflared-tunnel-setup/revert_broken.py

"""Revert the 3 hostnames that have no functional backend:
- plex (NPM has no vhost)
- rustdesk (NPM has no vhost)
- secure (Jupiter can't route to 172.16.1.16)

Removes them from tunnel ingress and restores their original A records.
"""
import json, os, subprocess, urllib.error, urllib.request, time
import paramiko, yaml

ZONE = '1beb9917c22b54be32e5215df2c227ce'
CF_TOKEN = os.environ.get('CF_API_TOKEN_FULL_DNS','')
if not CF_TOKEN: raise SystemExit('set CF_API_TOKEN_FULL_DNS')

REVERT = {
    # hostname: original A content
    'plex.azcomputerguru.com':     '72.194.62.4',
    'rustdesk.azcomputerguru.com': '72.194.62.10',
    'secure.azcomputerguru.com':   '72.194.62.2',
}

def cfapi(method, path, body=None):
    req = urllib.request.Request(
        f'https://api.cloudflare.com/client/v4{path}',
        data=json.dumps(body).encode() if body else None,
        method=method,
        headers={'Authorization': f'Bearer {CF_TOKEN}','Content-Type':'application/json'},
    )
    try:
        with urllib.request.urlopen(req, timeout=30) as r:
            return json.loads(r.read())
    except urllib.error.HTTPError as e:
        try: return json.loads(e.read())
        except: return {'success':False,'errors':[{'message':str(e)}]}

def _pwd(v): return yaml.safe_load(subprocess.run(['sops','-d',v],capture_output=True,text=True,timeout=30,check=True).stdout)['credentials']['password']

j = paramiko.SSHClient(); j.set_missing_host_key_policy(paramiko.AutoAddPolicy())
j.connect('172.16.3.20', username='root', password=_pwd('D:/vault/infrastructure/jupiter-unraid-primary.sops.yaml'),
          timeout=30, look_for_keys=False, allow_agent=False)

def jrun(cmd, to=60):
    _, o, _ = j.exec_command(cmd, timeout=to)
    return o.read().decode()

try:
    print('=== [1] rewrite config.yml without the 3 broken hosts ===')
    APPDATA = '/mnt/cache/appdata/cloudflared'
    # Read UUID
    UUID = jrun(f'grep "^tunnel:" {APPDATA}/config.yml').split(':',1)[1].strip()

    IX = 'https://172.16.3.10:443'
    JNPM = 'https://172.16.3.20:18443'
    KEEP = [
        ('azcomputerguru.com', IX),
        ('analytics.azcomputerguru.com', IX),
        ('community.azcomputerguru.com', IX),
        ('radio.azcomputerguru.com', IX),
        ('ix.azcomputerguru.com', IX),
        ('git.azcomputerguru.com', JNPM),
        ('plexrequest.azcomputerguru.com', JNPM),
        ('rmm.azcomputerguru.com', JNPM),
        ('rmm-api.azcomputerguru.com', JNPM),
        ('sync.azcomputerguru.com', JNPM),
    ]
    config = f'tunnel: {UUID}\ncredentials-file: /home/nonroot/.cloudflared/{UUID}.json\ningress:\n'
    for h, svc in KEEP:
        config += f'  - hostname: {h}\n    service: {svc}\n    originRequest:\n      originServerName: {h}\n      noTLSVerify: true\n'
    config += '  - service: http_status:404\n'
    jrun(f'cp {APPDATA}/config.yml {APPDATA}/config.yml.bak-$(date +%Y%m%d-%H%M%S)')
    HD = "'EOF_CFG'"
    jrun(f"cat > {APPDATA}/config.yml <<{HD}\n{config}\nEOF_CFG")
    jrun(f'chown 65532:65532 {APPDATA}/config.yml')
    print(f'  10 ingress hostnames kept (plex/rustdesk/secure removed)')

    print('\n=== [2] revert DNS for 3 hosts ===')
    for host, orig_ip in REVERT.items():
        r = cfapi('GET', f'/zones/{ZONE}/dns_records?name={host}')
        if not r.get('success') or not r['result']:
            print(f'  [{host}] no record, skipping'); continue
        rec = r['result'][0]
        print(f'  [{host}] current: type={rec["type"]} content={rec["content"]}')
        d = cfapi('DELETE', f'/zones/{ZONE}/dns_records/{rec["id"]}')
        if not d.get('success'):
            print(f'    [FAIL delete] {d.get("errors")}'); continue
        body = {'type':'A','name':host,'content':orig_ip,'proxied':True,'ttl':1}
        cr = cfapi('POST', f'/zones/{ZONE}/dns_records', body)
        if cr.get('success'):
            print(f'    [OK] restored A {orig_ip} proxied')
        else:
            print(f'    [FAIL create] {cr.get("errors")}')

    print('\n=== [3] restart cloudflared ===')
    print(jrun('docker restart cloudflared').rstrip())

    print('\n=== [4] wait for reconnect ===')
    for i in range(20):
        time.sleep(3)
        logs = jrun('docker logs cloudflared 2>&1 | tail -30')
        conns = logs.count('Registered tunnel connection')
        if conns >= 4:
            print(f'  [try {i+1}] {conns} connections')
            break
finally:
    j.close()

print('\n=== [5] external probe all 10 tunneled hostnames ===')
import urllib.request
for h in [k[0] for k in [
    ('azcomputerguru.com',),('analytics.azcomputerguru.com',),('community.azcomputerguru.com',),
    ('radio.azcomputerguru.com',),('ix.azcomputerguru.com',),('git.azcomputerguru.com',),
    ('plexrequest.azcomputerguru.com',),('rmm.azcomputerguru.com',),('rmm-api.azcomputerguru.com',),
    ('sync.azcomputerguru.com',),
]]:
    try:
        req = urllib.request.Request(f'https://{h}/', method='HEAD',
            headers={'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) Chrome/120.0'})
        with urllib.request.urlopen(req, timeout=15) as r:
            print(f'  {h:42} HTTP {r.status} {r.headers.get("Server","-")}')
    except urllib.error.HTTPError as e:
        print(f'  {h:42} HTTP {e.code}')
    except Exception as e:
        print(f'  {h:42} ERR {str(e)[:40]}')