Enforce single instance Local Agent supervisor worker and dashboard

2026-09-03 14:03:26 +05:30
parent 010d0969b4
commit 5dce7f16f9
7 changed files with 267 additions and 3 deletions
@@ -4,7 +4,7 @@ import io
 from pathlib import Path
 import zipfile

-ERP_LOCAL_AGENT_VERSION = "1.22.2"
+ERP_LOCAL_AGENT_VERSION = "1.22.3"
 ERP_LOCAL_AGENT_NAME = "ERP Local Agent"
 RUNTIME_ROOT = Path(__file__).resolve().parent / "local_agent_runtime"
 _DETERMINISTIC_ZIP_TIMESTAMP = (2026, 1, 1, 0, 0, 0)
@@ -48,6 +48,17 @@ def _build_zip(*, env_text: str | None, include_env: bool, include_admin_readme:
            if not path.is_file() or "__pycache__" in path.parts:
                continue
            _write_zip_bytes(dst, path.relative_to(RUNTIME_ROOT).as_posix(), path.read_bytes())
+        # Self-healing desktop launchers. Older supervisors update only the
+        # erp_local_agent package, so place canonical copies inside the worker
+        # package as data. The updated worker promotes them to INSTALL_ROOT.
+        for helper_name in ("ERPAgentSupervisor.pyw", "ERPAgentDashboard.pyw"):
+            helper = RUNTIME_ROOT / helper_name
+            if helper.is_file():
+                _write_zip_bytes(
+                    dst,
+                    f"erp_local_agent/_desktop_runtime/{helper_name}",
+                    helper.read_bytes(),
+                )
        if include_env and env_text is not None:
            _write_zip_bytes(dst, ".env", env_text.encode("utf-8"))
        if include_admin_readme:
@@ -1,6 +1,10 @@
 from __future__ import annotations

+import ctypes
+from ctypes import wintypes
+import hashlib
 import os
+import atexit
 from pathlib import Path
 import subprocess
 import sys
@@ -16,6 +20,63 @@ DATA_DIR = INSTALL_ROOT / "data"
 SUPERVISOR_LOCK = DATA_DIR / "supervisor.lock"
 SUPERVISOR_SCRIPT = INSTALL_ROOT / "ERPAgentSupervisor.pyw"
 URL = "http://127.0.0.1:8788"
+ERROR_ALREADY_EXISTS = 183
+
+
+def _dashboard_mutex_name() -> str:
+    normalized = str(INSTALL_ROOT.resolve()).casefold().encode("utf-8", errors="ignore")
+    suffix = hashlib.sha256(normalized).hexdigest()[:20]
+    return f"Local\\ARRR_ERP_Local_Agent_DASHBOARD_{suffix}"
+
+
+def _acquire_dashboard_mutex():
+    if os.name != "nt":
+        return object()
+    kernel32 = ctypes.WinDLL("kernel32", use_last_error=True)
+    kernel32.CreateMutexW.argtypes = (ctypes.c_void_p, wintypes.BOOL, wintypes.LPCWSTR)
+    kernel32.CreateMutexW.restype = wintypes.HANDLE
+    ctypes.set_last_error(0)
+    handle = kernel32.CreateMutexW(None, False, _dashboard_mutex_name())
+    if not handle:
+        raise ctypes.WinError(ctypes.get_last_error())
+    if ctypes.get_last_error() == ERROR_ALREADY_EXISTS:
+        kernel32.CloseHandle(handle)
+        return None
+    return handle
+
+
+def _close_dashboard_mutex(handle) -> None:
+    if os.name != "nt" or handle is None or not isinstance(handle, int):
+        return
+    try:
+        ctypes.WinDLL("kernel32", use_last_error=True).CloseHandle(handle)
+    except Exception:
+        pass
+
+
+def _focus_existing_dashboard() -> None:
+    if os.name != "nt":
+        return
+    try:
+        user32 = ctypes.WinDLL("user32", use_last_error=True)
+        EnumWindowsProc = ctypes.WINFUNCTYPE(wintypes.BOOL, wintypes.HWND, wintypes.LPARAM)
+
+        @EnumWindowsProc
+        def callback(hwnd, _lparam):
+            length = user32.GetWindowTextLengthW(hwnd)
+            if length <= 0:
+                return True
+            buf = ctypes.create_unicode_buffer(length + 1)
+            user32.GetWindowTextW(hwnd, buf, length + 1)
+            if "ERP Local Agent Dashboard" in buf.value:
+                user32.ShowWindow(hwnd, 9)  # SW_RESTORE
+                user32.SetForegroundWindow(hwnd)
+                return False
+            return True
+
+        user32.EnumWindows(callback, 0)
+    except Exception:
+        pass


 def _create_no_window() -> int:
@@ -115,6 +176,11 @@ def _wait_ready(seconds: int = 75) -> bool:


 def main() -> None:
+    dashboard_mutex = _acquire_dashboard_mutex()
+    if dashboard_mutex is None:
+        _focus_existing_dashboard()
+        return
+    atexit.register(_close_dashboard_mutex, dashboard_mutex)
    try:
        _ensure_supervisor()
    except Exception as exc:
@@ -1,6 +1,9 @@
 from __future__ import annotations

 import argparse
+import ctypes
+from ctypes import wintypes
+import hashlib
 import json
 import os
 from pathlib import Path
@@ -25,6 +28,60 @@ DASHBOARD_URL = "http://127.0.0.1:8788"
 WORKER_MODULE = "erp_local_agent.main"
 RESTART_DELAY_SECONDS = 3
 HEALTH_TIMEOUT_SECONDS = 60
+ERROR_ALREADY_EXISTS = 183
+
+
+def _supervisor_mutex_name() -> str:
+    normalized = str(INSTALL_ROOT.resolve()).casefold().encode("utf-8", errors="ignore")
+    suffix = hashlib.sha256(normalized).hexdigest()[:20]
+    return f"Local\\ARRR_ERP_Local_Agent_SUPERVISOR_{suffix}"
+
+
+def _acquire_os_mutex():
+    if os.name != "nt":
+        return object()
+    kernel32 = ctypes.WinDLL("kernel32", use_last_error=True)
+    kernel32.CreateMutexW.argtypes = (ctypes.c_void_p, wintypes.BOOL, wintypes.LPCWSTR)
+    kernel32.CreateMutexW.restype = wintypes.HANDLE
+    ctypes.set_last_error(0)
+    handle = kernel32.CreateMutexW(None, False, _supervisor_mutex_name())
+    if not handle:
+        raise ctypes.WinError(ctypes.get_last_error())
+    if ctypes.get_last_error() == ERROR_ALREADY_EXISTS:
+        kernel32.CloseHandle(handle)
+        return None
+    return handle
+
+
+def _close_os_mutex(handle) -> None:
+    if os.name != "nt" or handle is None or not isinstance(handle, int):
+        return
+    try:
+        ctypes.WinDLL("kernel32", use_last_error=True).CloseHandle(handle)
+    except Exception:
+        pass
+
+
+def _kill_duplicate_supervisors() -> None:
+    if os.name != "nt":
+        return
+    root = str(INSTALL_ROOT).replace("'", "''")
+    script = (
+        "$root='" + root + "';$self=" + str(os.getpid()) + ";"
+        "Get-CimInstance Win32_Process -ErrorAction SilentlyContinue | "
+        "Where-Object { $_.ProcessId -ne $self -and $_.CommandLine -and "
+        "$_.CommandLine -match 'ERPAgentSupervisor\\.pyw' -and "
+        "$_.CommandLine -like ('*'+$root+'*') } | "
+        "ForEach-Object { Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue }"
+    )
+    subprocess.run(
+        ["powershell.exe", "-NoProfile", "-ExecutionPolicy", "Bypass", "-Command", script],
+        cwd=str(INSTALL_ROOT),
+        stdout=subprocess.DEVNULL,
+        stderr=subprocess.DEVNULL,
+        creationflags=_create_no_window(),
+        check=False,
+    )


 def _create_no_window() -> int:
@@ -135,8 +192,14 @@ class Supervisor:
    def __init__(self):
        self.worker: subprocess.Popen | None = None
        self.running = True
+        self._os_mutex = None

    def acquire_lock(self) -> bool:
+        # The named mutex is atomic at the Windows kernel level and prevents
+        # the lock-file race that allowed multiple supervisors before 1.22.3.
+        self._os_mutex = _acquire_os_mutex()
+        if self._os_mutex is None:
+            return False
        DATA_DIR.mkdir(parents=True, exist_ok=True)
        if LOCK_FILE.exists():
            try:
@@ -154,6 +217,8 @@ class Supervisor:
                LOCK_FILE.unlink()
        except Exception:
            pass
+        _close_os_mutex(self._os_mutex)
+        self._os_mutex = None

    def start_worker(self) -> None:
        if self.worker is not None and self.worker.poll() is None:
@@ -368,6 +433,7 @@ class Supervisor:
            return 0

        try:
+            _kill_duplicate_supervisors()
            _state("starting", "ERP Local Agent supervisor starting.")
            self.start_worker()

@@ -1,2 +1,2 @@
-__version__ = "1.22.2"
+__version__ = "1.22.3"
 AGENT_NAME = "ERP Local Agent"
@@ -0,0 +1,107 @@
+from __future__ import annotations
+
+import ctypes
+from ctypes import wintypes
+import hashlib
+import os
+from pathlib import Path
+import subprocess
+
+
+ERROR_ALREADY_EXISTS = 183
+
+
+def _mutex_name(scope: str, root: Path) -> str:
+    normalized = str(root.resolve()).casefold().encode("utf-8", errors="ignore")
+    suffix = hashlib.sha256(normalized).hexdigest()[:20]
+    return f"Local\\ARRR_ERP_Local_Agent_{scope}_{suffix}"
+
+
+def acquire_named_mutex(scope: str, root: Path):
+    """Return an owned Windows mutex handle, or None when another instance owns it."""
+    if os.name != "nt":
+        return object()
+    kernel32 = ctypes.WinDLL("kernel32", use_last_error=True)
+    kernel32.CreateMutexW.argtypes = (ctypes.c_void_p, wintypes.BOOL, wintypes.LPCWSTR)
+    kernel32.CreateMutexW.restype = wintypes.HANDLE
+    ctypes.set_last_error(0)
+    handle = kernel32.CreateMutexW(None, False, _mutex_name(scope, root))
+    if not handle:
+        raise ctypes.WinError(ctypes.get_last_error())
+    if ctypes.get_last_error() == ERROR_ALREADY_EXISTS:
+        kernel32.CloseHandle(handle)
+        return None
+    return handle
+
+
+def close_named_mutex(handle) -> None:
+    if os.name != "nt" or handle is None or not isinstance(handle, int):
+        return
+    try:
+        ctypes.WinDLL("kernel32", use_last_error=True).CloseHandle(handle)
+    except Exception:
+        pass
+
+
+def sync_desktop_helpers(root: Path, logger=None) -> bool:
+    """Promote canonical desktop launchers shipped inside the worker package."""
+    source_root = Path(__file__).resolve().parent / "_desktop_runtime"
+    changed = False
+    if not source_root.is_dir():
+        return False
+    for name in ("ERPAgentSupervisor.pyw", "ERPAgentDashboard.pyw"):
+        source = source_root / name
+        target = root / name
+        if not source.is_file():
+            continue
+        source_bytes = source.read_bytes()
+        try:
+            if target.is_file() and target.read_bytes() == source_bytes:
+                continue
+        except Exception:
+            pass
+        tmp = target.with_suffix(target.suffix + ".new")
+        tmp.write_bytes(source_bytes)
+        os.replace(tmp, target)
+        changed = True
+        if logger:
+            logger.info("Refreshed Local Agent desktop helper: %s", name)
+    return changed
+
+
+def _run_powershell(script: str) -> None:
+    if os.name != "nt":
+        return
+    subprocess.run(
+        ["powershell.exe", "-NoProfile", "-ExecutionPolicy", "Bypass", "-Command", script],
+        stdout=subprocess.DEVNULL,
+        stderr=subprocess.DEVNULL,
+        creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0),
+        check=False,
+        timeout=30,
+    )
+
+
+def cleanup_legacy_duplicates(root: Path, *, keep_supervisor_pid: int, close_dashboards: bool) -> None:
+    """Remove duplicates left behind by pre-1.22.3 launchers on this installation only."""
+    if os.name != "nt":
+        return
+    root_text = str(root.resolve()).replace("'", "''")
+    dashboard_clause = ""
+    if close_dashboards:
+        dashboard_clause = " -or ($cmd -match 'ERPAgentDashboard\\.pyw')"
+    script = rf"""
+$root='{root_text}'
+$keepSupervisor={int(keep_supervisor_pid or 0)}
+$selfPid={os.getpid()}
+Get-CimInstance Win32_Process -ErrorAction SilentlyContinue | ForEach-Object {{
+    $cmd = [string]$_.CommandLine
+    if (-not $cmd -or $cmd -notlike ('*'+$root+'*')) {{ return }}
+    $kill = $false
+    if (($cmd -match 'ERPAgentSupervisor\.pyw') -and ($_.ProcessId -ne $keepSupervisor)) {{ $kill = $true }}
+    if (($cmd -match 'erp_local_agent\.main') -and ($_.ProcessId -ne $selfPid)) {{ $kill = $true }}
+    if (($cmd -match 'ERPAgentDashboard\.pyw') -and {str(bool(close_dashboards)).lower()}) {{ $kill = $true }}
+    if ($kill) {{ Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue }}
+}}
+"""
+    _run_powershell(script)
@@ -2,6 +2,7 @@ from __future__ import annotations

 import argparse
 import asyncio
+import os
 from dataclasses import replace
 from pathlib import Path
 import threading
@@ -12,6 +13,7 @@ from .client import ERPClient
 from .config import load_config
 from .dashboard import AgentDashboard
 from .db import LocalDB
+from .desktop_runtime import acquire_named_mutex, cleanup_legacy_duplicates, close_named_mutex, sync_desktop_helpers
 from .logger import setup_logger
 from .sync import StorageAgent
 from .tunnel import StorageAgentTunnel
@@ -29,9 +31,19 @@ def build_parser() -> argparse.ArgumentParser:
 def main() -> int:
    args = build_parser().parse_args()
    root = Path.cwd()
+    worker_mutex = acquire_named_mutex("WORKER", root)
+    if worker_mutex is None:
+        # A supervisor already owns the one permitted worker for this install.
+        return 0
    logger = setup_logger(root)
    db = None
    try:
+        helpers_changed = sync_desktop_helpers(root, logger=logger)
+        cleanup_legacy_duplicates(
+            root,
+            keep_supervisor_pid=os.getppid(),
+            close_dashboards=helpers_changed,
+        )
        config = load_config(args.env)
        db = LocalDB(root / "data" / "agent.db")
        db.prune_history()
@@ -84,6 +96,8 @@ def main() -> int:
            except Exception:
                pass
        return 1
+    finally:
+        close_named_mutex(worker_mutex)


 if __name__ == "__main__":
@@ -13,7 +13,7 @@ set "AGENT_DIR=%~dp0"
 set "AGENT_PYTHONW=%~dp0.venv\Scripts\pythonw.exe"
 set "SUPERVISOR=%~dp0ERPAgentSupervisor.pyw"

-powershell.exe -NoProfile -ExecutionPolicy Bypass -Command "$ErrorActionPreference='Stop'; try { Stop-ScheduledTask -TaskName 'AuditFirmStorageAgent' -ErrorAction SilentlyContinue } catch {}; try { Unregister-ScheduledTask -TaskName 'AuditFirmStorageAgent' -Confirm:$false -ErrorAction SilentlyContinue } catch {}; try { Stop-ScheduledTask -TaskName 'ERP Local Agent' -ErrorAction SilentlyContinue } catch {}; try { Unregister-ScheduledTask -TaskName 'ERP Local Agent' -Confirm:$false -ErrorAction SilentlyContinue } catch {}; Get-CimInstance Win32_Process -ErrorAction SilentlyContinue | Where-Object { $_.CommandLine -and (($_.CommandLine -match 'ERPAgentSupervisor\.pyw') -or ($_.CommandLine -match 'erp_local_agent\.main')) -and $_.CommandLine -like ('*'+$env:AGENT_DIR.TrimEnd('\')+'*') } | ForEach-Object { Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue }; $d=$env:AGENT_DIR.TrimEnd('\'); $p=$env:AGENT_PYTHONW; $s=$env:SUPERVISOR; $a=New-ScheduledTaskAction -Execute $p -Argument ('\"'+$s+'\" --background') -WorkingDirectory $d; $t=New-ScheduledTaskTrigger -AtStartup; $pr=New-ScheduledTaskPrincipal -UserId 'SYSTEM' -LogonType ServiceAccount -RunLevel Highest; $settings=New-ScheduledTaskSettingsSet -RestartCount 999 -RestartInterval (New-TimeSpan -Minutes 1) -ExecutionTimeLimit ([TimeSpan]::Zero) -AllowStartIfOnBatteries -DontStopIfGoingOnBatteries; Register-ScheduledTask -TaskName 'ERP Local Agent' -Action $a -Trigger $t -Principal $pr -Settings $settings -Force | Out-Null; Start-ScheduledTask -TaskName 'ERP Local Agent'"
+powershell.exe -NoProfile -ExecutionPolicy Bypass -Command "$ErrorActionPreference='Stop'; try { Stop-ScheduledTask -TaskName 'AuditFirmStorageAgent' -ErrorAction SilentlyContinue } catch {}; try { Unregister-ScheduledTask -TaskName 'AuditFirmStorageAgent' -Confirm:$false -ErrorAction SilentlyContinue } catch {}; try { Stop-ScheduledTask -TaskName 'ERP Local Agent' -ErrorAction SilentlyContinue } catch {}; try { Unregister-ScheduledTask -TaskName 'ERP Local Agent' -Confirm:$false -ErrorAction SilentlyContinue } catch {}; Get-CimInstance Win32_Process -ErrorAction SilentlyContinue | Where-Object { $_.CommandLine -and (($_.CommandLine -match 'ERPAgentSupervisor\.pyw') -or ($_.CommandLine -match 'erp_local_agent\.main')) -and $_.CommandLine -like ('*'+$env:AGENT_DIR.TrimEnd('\')+'*') } | ForEach-Object { Stop-Process -Id $_.ProcessId -Force -ErrorAction SilentlyContinue }; $d=$env:AGENT_DIR.TrimEnd('\'); $p=$env:AGENT_PYTHONW; $s=$env:SUPERVISOR; $a=New-ScheduledTaskAction -Execute $p -Argument ('\"'+$s+'\" --background') -WorkingDirectory $d; $t=New-ScheduledTaskTrigger -AtStartup; $pr=New-ScheduledTaskPrincipal -UserId 'SYSTEM' -LogonType ServiceAccount -RunLevel Highest; $settings=New-ScheduledTaskSettingsSet -RestartCount 999 -RestartInterval (New-TimeSpan -Minutes 1) -ExecutionTimeLimit ([TimeSpan]::Zero) -AllowStartIfOnBatteries -DontStopIfGoingOnBatteries -MultipleInstances IgnoreNew; Register-ScheduledTask -TaskName 'ERP Local Agent' -Action $a -Trigger $t -Principal $pr -Settings $settings -Force | Out-Null; Start-ScheduledTask -TaskName 'ERP Local Agent'"
 if errorlevel 1 goto :failed

 echo ERP Local Agent Supervisor installed and started successfully.