Files
agent-workbench/scripts/bridge_service.py

67 lines
3.3 KiB
Python

#!/usr/bin/env python3
"""Keep the local queue worker and visible-history mirror available without exporting auth."""
import argparse
import json
import os
from pathlib import Path
import subprocess
import sys
import time
from worker import atomic_json,session_lock
from faults import report_process
def main():
parser=argparse.ArgumentParser(description=__doc__)
parser.add_argument('--config',required=True)
args=parser.parse_args()
config_path=Path(args.config).resolve()
cfg=json.loads(config_path.read_text(encoding='utf-8-sig'))
root=Path(cfg['state_file']).expanduser().resolve().parent
root.mkdir(parents=True,exist_ok=True,mode=0o700)
if os.name=='posix':os.umask(0o077)
identity='supervisor:'+json.dumps({'control':cfg['control'],'session_id':cfg['session_id']},sort_keys=True)
fault_path=root/'bridge-faults.json'
faults=json.loads(fault_path.read_text(encoding='utf-8')).get('roles',{}) if fault_path.exists() else {}
processes={}
started_at={}
logs={}
next_start={}
with session_lock(identity):
try:
while True:
state_path=Path(cfg['state_file']).expanduser()
state=json.loads(state_path.read_text(encoding='utf-8')) if state_path.exists() else {}
for role in (('worker','mirror','notifier') if cfg.get('notification_targets') else ('worker','mirror')):
process=processes.get(role)
if process and cfg.get('notification_targets'):
report_process(cfg,root,role,process,faults,started_at.get(role))
if process and process.poll() is None:continue
if role=='worker' and (state.get('needs_reconciliation') or (state.get('inflight') and not state.get('pending_completion'))):
continue
if time.monotonic()<next_start.get(role,0):continue
if role not in logs:logs[role]=open(root/(role+'.log'),'a',encoding='utf-8')
command=[sys.executable,'-X','utf8',str(Path(__file__).with_name(role+'.py')),'--config',str(config_path)]
process=subprocess.Popen(command,stdout=logs[role],stderr=logs[role],
**({'creationflags':subprocess.CREATE_NO_WINDOW} if os.name=='nt' else {}))
processes[role]=process
started_at[role]=time.time()
next_start[role]=time.monotonic()+15
atomic_json(root/'bridge-service.json',{'pid':os.getpid(),'updated':time.time(),
'children':{name:{'pid':p.pid,'running':p.poll() is None} for name,p in processes.items()},
'worker_paused_for_reconciliation':bool((state.get('needs_reconciliation') or (state.get('inflight') and not state.get('pending_completion')))
and ('worker' not in processes or processes['worker'].poll() is not None))})
time.sleep(5)
except KeyboardInterrupt:
pass
finally:
for process in processes.values():
if process.poll() is None:process.terminate()
for process in processes.values():
try:process.wait(timeout=10)
except subprocess.TimeoutExpired:process.kill()
for stream in logs.values():stream.close()
if __name__=='__main__':main()