Add event-driven reporting and durable failure recovery hooks

scripts/bridge_service.py

@@ -8,6 +8,7 @@ import subprocess
 import sys
 import time
 from worker import atomic_json,session_lock
+from faults import report_process
 
 
 def main():
@@ -20,7 +21,10 @@ def main():
     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):
@@ -28,8 +32,10 @@ def main():
             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'):
+                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
@@ -39,6 +45,7 @@ def main():
                     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()},