-
Notifications
You must be signed in to change notification settings - Fork 17
Expand file tree
/
Copy pathcontrol_server.py
More file actions
225 lines (187 loc) · 7.52 KB
/
control_server.py
File metadata and controls
225 lines (187 loc) · 7.52 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
"""WebSyn control plane on :8101.
Endpoints:
GET /health -> per-site PID + alive status
POST /reset/<site> -> SIGTERM site, restore instance/ from instance_seed/, respawn
POST /reset-all -> reset every site in parallel
POST /restart/<site> -> just respawn (no DB wipe) -- bonus, useful for code reload
PID tracking: each site's PID lives at /tmp/websyn_pids/<site>.pid. websyn_start.sh
writes the initial PIDs; this server overwrites them on respawn.
"""
import os
import shutil
import signal
import subprocess
import sys
import threading
import time
import urllib.request
from concurrent.futures import ThreadPoolExecutor
from pathlib import Path
from flask import Flask, jsonify
SITES = [
'allrecipes', 'amazon', 'apple', 'arxiv', 'bbc_news', 'booking',
'github', 'google_flights', 'google_map', 'google_search',
'huggingface', 'wolfram_alpha', 'cambridge_dictionary',
'coursera', 'espn',
]
BASE_PORT = 40000
WEBSYN_DIR = '/opt/WebSyn'
PID_DIR = Path('/tmp/websyn_pids')
PID_DIR.mkdir(parents=True, exist_ok=True)
# Per-site mutex so concurrent /reset, /reset-all, /restart against the same
# site can't race on PID files / instance dir.
_site_locks = {s: threading.Lock() for s in SITES}
# Per-site Popen handles for the supervisor processes that control_server
# spawned itself. Keeping the Popen reference lets us .kill() + .wait() to
# both signal AND reap the process group cleanly. Without it, Python's
# garbage collector eventually wait()s, but until then is_alive(pid) keeps
# returning True for the zombie and our reset loop blocks on a stale poll.
#
# Sites started by websyn_start.sh at boot won't have an entry here until
# their first respawn — kill_site falls back to os.killpg + zombie poll
# for those.
_site_procs: dict = {}
# We tried graceful SIGTERM. Werkzeug's threaded serve_forever() doesn't
# honor it. Since /reset wipes instance/ next anyway, in-flight transactions
# are about to be discarded — graceful shutdown has no value here. SIGKILL
# directly via process group (site_runner.py runs setsid → its pgid == pid).
REAP_GRACE_SECS = 5.0
app = Flask(__name__)
def site_port(site: str) -> int:
return BASE_PORT + SITES.index(site)
def pid_path(site: str) -> Path:
return PID_DIR / f'{site}.pid'
def read_pid(site: str):
try:
return int(pid_path(site).read_text().strip())
except (FileNotFoundError, ValueError):
return None
def is_alive(pid) -> bool:
"""True iff a runnable / sleeping process with this PID exists.
Returns False for zombies and missing PIDs — that is what the poll
loops in kill_site need."""
if not pid:
return False
try:
with open(f'/proc/{pid}/stat', 'rb') as f:
data = f.read()
# /proc/<pid>/stat: "<pid> (<comm>) <state> ..." comm may contain
# spaces or parens, so split on the LAST ')'.
rparen = data.rindex(b')')
state = data[rparen + 2:rparen + 3]
return state not in (b'Z', b'X')
except (FileNotFoundError, ProcessLookupError, ValueError):
return False
def kill_site(site: str, reap_grace: float = REAP_GRACE_SECS):
pid = read_pid(site)
if not pid:
return
# SIGKILL the whole process group (supervisor + Flask child). Without
# killpg the Flask child would re-parent to init and keep the port.
try:
os.killpg(pid, signal.SIGKILL)
except ProcessLookupError:
pass
# If we own a Popen for this supervisor, wait()+reap so it doesn't
# linger as a zombie. (Supervisors started at boot via websyn_start.sh
# aren't tracked here; we still adopted them as children via container
# init, but Python won't reap them — they stay zombies until container
# exit. That's harmless: is_alive() correctly reports them as dead.)
proc = _site_procs.pop(site, None)
if proc is not None:
try:
proc.wait(timeout=reap_grace)
except subprocess.TimeoutExpired:
pass
# Belt-and-suspenders: confirm the supervisor is actually dead before
# returning, even when we don't own the Popen. is_alive() looks at
# /proc state and returns False for zombies, so this loop exits in ms.
deadline = time.time() + reap_grace
while time.time() < deadline:
if not is_alive(pid):
return
time.sleep(0.01)
def reset_db(site: str):
inst = Path(WEBSYN_DIR) / site / 'instance'
seed = Path(WEBSYN_DIR) / site / 'instance_seed'
if inst.exists():
shutil.rmtree(inst)
shutil.copytree(seed, inst)
def start_site(site: str) -> int:
port = site_port(site)
log = open(f'/tmp/websyn_{site}.log', 'a', buffering=1)
log.write(f'\n[control-server] respawn at {time.strftime("%F %T")}\n')
# Run Flask under /opt/site_runner.py supervisor — see that file for
# the rationale. start_new_session=True is redundant with the supervisor's
# own setsid() but harmless and gives us a session leader from the very
# first instant.
proc = subprocess.Popen(
['python3', '/opt/site_runner.py', site, str(port)],
stdout=log, stderr=subprocess.STDOUT,
start_new_session=True,
)
_site_procs[site] = proc
pid_path(site).write_text(str(proc.pid))
return proc.pid
def wait_ready(site: str, timeout: float = 30.0) -> bool:
port = site_port(site)
deadline = time.time() + timeout
while time.time() < deadline:
try:
urllib.request.urlopen(f'http://127.0.0.1:{port}/', timeout=2).read(1)
return True
except Exception:
time.sleep(0.3)
return False
def reset_one(site: str) -> dict:
with _site_locks[site]:
kill_site(site)
reset_db(site)
pid = start_site(site)
ready = wait_ready(site)
return {'site': site, 'pid': pid, 'ready': ready}
def restart_one(site: str) -> dict:
"""Just respawn — no DB wipe."""
with _site_locks[site]:
kill_site(site)
pid = start_site(site)
ready = wait_ready(site)
return {'site': site, 'pid': pid, 'ready': ready,
'note': 'restart only, DB not reset'}
@app.route('/health')
def health():
sites = {}
all_ok = True
for s in SITES:
pid = read_pid(s)
alive = is_alive(pid)
sites[s] = {'pid': pid, 'alive': alive, 'port': site_port(s)}
if not alive:
all_ok = False
return jsonify({'ok': all_ok, 'sites': sites}), (200 if all_ok else 503)
@app.route('/reset/<site>', methods=['POST'])
def reset_site(site):
if site not in SITES:
return jsonify({'error': f'unknown site: {site}',
'valid_sites': SITES}), 404
result = reset_one(site)
return jsonify(result), (200 if result['ready'] else 503)
@app.route('/reset-all', methods=['POST'])
def reset_all():
with ThreadPoolExecutor(max_workers=len(SITES)) as ex:
results = list(ex.map(reset_one, SITES))
out = {r['site']: r for r in results}
ok = all(r['ready'] for r in results)
return jsonify({'ok': ok, 'sites': out}), (200 if ok else 503)
@app.route('/restart/<site>', methods=['POST'])
def restart_site(site):
if site not in SITES:
return jsonify({'error': f'unknown site: {site}'}), 404
result = restart_one(site)
return jsonify(result), (200 if result['ready'] else 503)
if __name__ == '__main__':
port = int(os.environ.get('CONTROL_PORT', 8101))
if '--port' in sys.argv:
port = int(sys.argv[sys.argv.index('--port') + 1])
app.run(host='0.0.0.0', port=port, debug=False,
use_reloader=False, threaded=True)