Files
security-alert-center/backend/app/services/host_remote_actions.py
T
PapaTramp 9883e6aab3 feat: per-host WinRM/SSH credentials override (0.5.17)
Allow host-level login/password for home/workgroup PCs while keeping global domain admins in Settings.
2026-07-14 20:04:42 +10:00

450 lines
15 KiB
Python

"""Background SSH/WinRM host actions (survives UI navigation and multi-worker API)."""
from __future__ import annotations
import logging
import os
import threading
from datetime import datetime, timezone
from typing import Any, Callable
from sqlalchemy.orm import Session
from app.database import SessionLocal
from app.models import Host
from app.services.agent_update import execute_agent_update_fallback
from app.services.agent_update_types import AgentUpdateFallbackResult
from app.services.linux_admin_settings import get_effective_linux_admin_for_host
from app.services.agent_update_settings import get_effective_agent_update_config
from app.services.ssh_connect import (
iter_ssh_targets,
run_ssh_monitor_update,
SshCommandResult,
tail_ssh_monitor_update_log,
)
logger = logging.getLogger(__name__)
ActionRunner = Callable[[Session, Host], AgentUpdateFallbackResult | SshCommandResult]
_lock = threading.Lock()
_running: set[int] = set()
class RemoteActionAlreadyRunningError(Exception):
pass
def _run_inline() -> bool:
return os.environ.get("SAC_REMOTE_ACTION_INLINE", "").strip().lower() in ("1", "true", "yes")
def _utcnow() -> datetime:
return datetime.now(timezone.utc)
def _format_output(result: AgentUpdateFallbackResult | SshCommandResult) -> str:
parts: list[str] = []
stdout = (getattr(result, "stdout", None) or "").strip()
stderr = (getattr(result, "stderr", None) or "").strip()
if stdout:
parts.append(stdout)
if stderr:
parts.append(stderr)
exit_code = getattr(result, "exit_code", None)
if exit_code is not None:
parts.append(f"exit code: {exit_code}")
return "\n\n".join(parts)
def _result_payload(
result: AgentUpdateFallbackResult | SshCommandResult,
*,
title: str,
started_at: str,
) -> dict[str, Any]:
finished = _utcnow().isoformat()
product_version = getattr(result, "product_version", None) or getattr(result, "agent_version", None)
return {
"title": title,
"status": "success" if result.ok else "failed",
"message": result.message,
"stdout": getattr(result, "stdout", None) or "",
"stderr": getattr(result, "stderr", None) or "",
"output": _format_output(result),
"ok": result.ok,
"exit_code": getattr(result, "exit_code", None),
"product_version": product_version,
"target": getattr(result, "target", None) or "",
"started_at": started_at,
"finished_at": finished,
}
def _completion_title(base_title: str, result: SshCommandResult) -> str:
if result.ok:
return f"{base_title} — готово"
return f"{base_title} — ошибка"
def _completion_message(result: SshCommandResult) -> str:
if result.ok:
version = (result.agent_version or "").strip()
if version:
return f"Обновление завершено успешно. Версия агента: {version}"
return "Обновление завершено успешно"
return (result.message or "Обновление не удалось").strip()
def _fetch_ssh_update_log_tail(db: Session, host: Host, *, lines: int = 200) -> str:
cfg = get_effective_linux_admin_for_host(db, host)
if not cfg.configured:
return ""
try:
targets = iter_ssh_targets(host)
except Exception:
return ""
for target in targets:
try:
tail = tail_ssh_monitor_update_log(
target=target,
user=cfg.user,
password=cfg.password,
lines=lines,
)
if tail:
return tail
except Exception:
logger.debug("final update log tail failed host_id=%s target=%s", host.id, target, exc_info=True)
return ""
def _enrich_ssh_update_payload(
payload: dict[str, Any],
*,
base_title: str,
result: SshCommandResult,
previous_output: str,
log_tail: str,
) -> dict[str, Any]:
payload["title"] = _completion_title(base_title, result)
payload["message"] = _completion_message(result)
if log_tail:
payload["output"] = log_tail
elif previous_output.strip():
payload["output"] = previous_output
return payload
def _mark_running(db: Session, host: Host, *, title: str) -> str:
started_at = _utcnow().isoformat()
host.agent_update_state = "running"
host.agent_update_last_error = None
host.remote_action = {
"title": title,
"status": "running",
"message": "Подключение к хосту и выполнение команды…",
"stdout": "",
"stderr": "",
"output": "",
"ok": None,
"exit_code": None,
"product_version": None,
"target": host.hostname,
"started_at": started_at,
"finished_at": None,
}
db.flush()
return started_at
def _apply_ssh_update_result(db: Session, host: Host, result: SshCommandResult) -> None:
now = _utcnow()
host.agent_update_last_at = now
if result.ok:
host.agent_update_state = "success"
host.agent_update_last_error = None
if result.agent_version:
host.product_version = result.agent_version
host.ssh_admin_ok = True
host.ssh_admin_checked_at = now
else:
host.agent_update_state = "failed"
host.agent_update_last_error = (result.message or "SSH update failed")[:2000]
_REMOTE_LOG_POLL_SEC = 2.0
def _poll_ssh_update_log_loop(
host_id: int,
*,
stop: threading.Event,
) -> None:
"""Периодически подтягивает tail update_script.log в hosts.remote_action для UI."""
while not stop.wait(_REMOTE_LOG_POLL_SEC):
session = SessionLocal()
try:
row = session.get(Host, host_id)
if row is None or (row.agent_update_state or "").strip().lower() != "running":
continue
payload = dict(row.remote_action or {})
if (payload.get("status") or "").strip().lower() != "running":
continue
cfg = get_effective_linux_admin_for_host(session, row)
if not cfg.configured:
continue
try:
targets = iter_ssh_targets(row)
except Exception:
continue
tail = ""
for target in targets:
try:
tail = tail_ssh_monitor_update_log(
target=target,
user=cfg.user,
password=cfg.password,
lines=200,
)
except Exception:
logger.debug("update log tail failed host_id=%s target=%s", host_id, target, exc_info=True)
continue
if tail:
break
session.refresh(row)
if (row.agent_update_state or "").strip().lower() != "running":
continue
payload = dict(row.remote_action or {})
if (payload.get("status") or "").strip().lower() != "running":
continue
if tail:
payload["output"] = tail
payload["message"] = "Выполняется обновление… (лог с хоста)"
row.remote_action = payload
session.commit()
except Exception:
logger.debug("remote log poll failed host_id=%s", host_id, exc_info=True)
session.rollback()
finally:
session.close()
def start_host_remote_action(
db: Session,
host: Host,
*,
title: str,
runner: ActionRunner,
poll_remote_log: bool = False,
) -> dict[str, Any]:
host_id = int(host.id)
with _lock:
if host_id in _running or host.agent_update_state == "running":
raise RemoteActionAlreadyRunningError(
f"Remote action already running for host {host.hostname}"
)
_running.add(host_id)
started_at = _mark_running(db, host, title=title)
db.commit()
def _worker() -> None:
session = SessionLocal()
stop_poll = threading.Event()
poll_thread: threading.Thread | None = None
if poll_remote_log and runner is run_ssh_monitor_update_action:
poll_thread = threading.Thread(
target=_poll_ssh_update_log_loop,
args=(host_id,),
kwargs={"stop": stop_poll},
name=f"sac-remote-log-{host_id}",
daemon=True,
)
poll_thread.start()
try:
row = session.get(Host, host_id)
if row is None:
return
previous_output = (row.remote_action or {}).get("output") or ""
result = runner(session, row)
started = (row.remote_action or {}).get("started_at") or started_at
payload = _result_payload(result, title=title, started_at=started)
if poll_remote_log and isinstance(result, SshCommandResult):
log_tail = _fetch_ssh_update_log_tail(session, row)
payload = _enrich_ssh_update_payload(
payload,
base_title=title,
result=result,
previous_output=previous_output,
log_tail=log_tail,
)
row.remote_action = payload
if isinstance(result, SshCommandResult):
_apply_ssh_update_result(session, row, result)
session.commit()
except Exception:
logger.exception("Background remote action failed for host_id=%s", host_id)
session.rollback()
row = session.get(Host, host_id)
if row is not None:
started = (row.remote_action or {}).get("started_at") or started_at
row.agent_update_state = "failed"
row.agent_update_last_error = "Internal error during remote action"
row.remote_action = {
"title": title,
"status": "failed",
"message": "Внутренняя ошибка SAC при выполнении действия",
"stdout": "",
"stderr": "",
"output": "",
"ok": False,
"exit_code": None,
"product_version": None,
"target": row.hostname,
"started_at": started,
"finished_at": _utcnow().isoformat(),
}
session.commit()
finally:
stop_poll.set()
if poll_thread is not None:
poll_thread.join(timeout=5.0)
session.close()
with _lock:
_running.discard(host_id)
if _run_inline():
_worker()
else:
threading.Thread(target=_worker, name=f"sac-remote-action-{host_id}", daemon=True).start()
return dict(host.remote_action or {})
def run_ssh_monitor_update_action(db: Session, host: Host) -> SshCommandResult:
cfg = get_effective_linux_admin_for_host(db, host)
if not cfg.configured:
return SshCommandResult(
ok=False,
message="Linux SSH admin is not configured",
target=host.hostname or "",
)
agent_cfg = get_effective_agent_update_config(db)
repo_url = (agent_cfg.ssh_git_repo_url or "").strip() or None
branch = (agent_cfg.git_branch or "main").strip() or "main"
try:
targets = iter_ssh_targets(host)
except Exception as exc:
return SshCommandResult(
ok=False,
message=str(exc),
target=host.hostname or "",
)
last: SshCommandResult | None = None
for target in targets:
last = run_ssh_monitor_update(
target=target,
user=cfg.user,
password=cfg.password,
repo_url=repo_url,
git_branch=branch,
)
if last.ok:
return last
return last or SshCommandResult(
ok=False,
message="SSH update failed",
target=host.hostname or "",
)
def run_agent_fallback_action(db: Session, host: Host) -> AgentUpdateFallbackResult:
return execute_agent_update_fallback(db, host)
def clear_stale_running_remote_actions(
db: Session,
*,
reason: str = "Прервано перезапуском SAC (фоновый worker не завершил job)",
host_ids: list[int] | None = None,
) -> list[str]:
"""Сбросить зависшие running после restart sac-api (daemon thread умер, запись в БД осталась)."""
from sqlalchemy import select
stmt = select(Host).where(Host.agent_update_state == "running")
if host_ids is not None:
stmt = stmt.where(Host.id.in_(host_ids))
rows = list(db.scalars(stmt).all())
if not rows:
return []
finished = _utcnow().isoformat()
hostnames: list[str] = []
for host in rows:
started = (host.remote_action or {}).get("started_at")
payload = dict(host.remote_action or {})
payload.update(
{
"title": payload.get("title") or "Remote action",
"status": "failed",
"message": reason,
"ok": False,
"finished_at": finished,
"started_at": started,
"target": payload.get("target") or host.hostname,
}
)
host.remote_action = payload
host.agent_update_state = "failed"
host.agent_update_last_error = reason[:2000]
hostnames.append(host.hostname or str(host.id))
db.commit()
with _lock:
for host in rows:
_running.discard(int(host.id))
return hostnames
def cancel_host_remote_action(db: Session, host: Host, *, reason: str | None = None) -> bool:
"""Отменить зависший running job вручную (admin)."""
if host.agent_update_state != "running":
return False
msg = reason or "Отменено администратором SAC"
cleared = clear_stale_running_remote_actions(db, reason=msg, host_ids=[int(host.id)])
return bool(cleared)
def get_remote_action_status(host: Host) -> dict[str, Any]:
payload = dict(host.remote_action or {})
if not payload:
return {"active": False, "host_id": host.id}
agent_state = (host.agent_update_state or "").strip().lower()
if agent_state == "running":
active = True
status = payload.get("status") or "running"
elif agent_state in ("success", "failed"):
active = False
status = payload.get("status") or agent_state
else:
status = payload.get("status") or host.agent_update_state or "unknown"
active = status == "running"
return {
"active": active,
"host_id": host.id,
"hostname": host.hostname,
"status": status,
"title": payload.get("title") or "",
"message": payload.get("message") or "",
"stdout": payload.get("stdout") or "",
"stderr": payload.get("stderr") or "",
"output": payload.get("output") or "",
"ok": payload.get("ok"),
"exit_code": payload.get("exit_code"),
"product_version": payload.get("product_version") or host.product_version,
"target": payload.get("target") or host.hostname,
"agent_update_state": host.agent_update_state,
"started_at": payload.get("started_at"),
"finished_at": payload.get("finished_at"),
}