Compare commits

...
3 Commits
Author SHA1 Message Date
lianlonggangandClaude Code 38a3a3ec3e fix: 看门狗拉起后增加宽限期防重启循环
主程序重启后 JVM 启动 + DB 建连可能耗时数分钟,300 秒宽限期内探测失败只记录不累计,避免重启循环。

Co-Authored-By: Claude Code <noreply@anthropic.com>
2026-09-09 14:32:30 +08:00
lianlonggangandClaude Code a6a801220c refactor: 移除日志实时流功能
删除 log_stream 模块及 LOG_TAIL_CTRL 命令、懒加载与停机释放逻辑,日志查看仅保留 LOG_TAIL 与 GET_LOG_DOWNLOAD。

Co-Authored-By: Claude Code <noreply@anthropic.com>
2026-09-09 14:32:29 +08:00
lianlonggangandClaude Code f7fbc054df feat: 升级下载 zip 后解压还原 jar
上传端为绕过中间设备对 jar 的拦截,升级包以 zip 下发;Agent 下载 zip 并 sha256 校验后,解压取内层 jar 再走替换/重启/回滚。

Co-Authored-By: Claude Code <noreply@anthropic.com>
2026-09-09 14:32:29 +08:00
5 changed files with 47 additions and 165 deletions
+1 -20
View File
@@ -9,7 +9,7 @@ B 端通过 WebSocket 下发 JSON 命令:{"type":"CMD_TYPE","payload":{...}}
- GET_VERSION 查询主程序版本
- GET_LOG_FILES 日志文件列表
- LOG_TAIL 日志尾部(payload: file, lines
- LOG_TAIL_CTRL 日志实时流开关payload: action, file
- GET_LOG_DOWNLOAD 日志全量下载payload: file
- RESTART 重启主程序(优雅停机后由 ProcessGuard 拉起)
- GET_CONFIG 读取医院配置文件(payload: hospital
- SAVE_CONFIG 写回医院配置文件(payload: hospital, content
@@ -36,8 +36,6 @@ class CommandDispatcher:
self._client = None # 由 main 注入(避免构造环)
self._upgrade_command = None # 升级命令(懒加载)
self._log_stream = None # 日志实时流管理(懒加载)
self._log_stream_lock = threading.Lock()
self._upgrade_lock = threading.Lock()
def set_client(self, client):
@@ -52,18 +50,6 @@ class CommandDispatcher:
self._upgrade_command = UpgradeCommand(self.config, self.main_app_client, self._client)
return self._upgrade_command
def _get_log_stream(self):
"""懒加载日志流管理器。"""
with self._log_stream_lock:
if self._log_stream is None:
from agent.log_stream import LogStreamManager
self._log_stream = LogStreamManager(self.config, self.main_app_client, self._client)
return self._log_stream
def shutdown_log_stream(self):
"""日志实时流(供 main 停机时统一释放)。"""
self._get_log_stream().stop_all()
# ================= 分发 =================
def dispatch(self, msg_type, payload):
@@ -89,11 +75,6 @@ class CommandDispatcher:
lines = (payload or {}).get("lines") or 200
result = {"ok": True, "data": self.main_app_client.log_tail(file, lines)}
elif msg_type == "LOG_TAIL_CTRL":
action = (payload or {}).get("action") or "start"
file = (payload or {}).get("file") or "info.log"
result = self._get_log_stream().control(action, file)
elif msg_type == "RESTART":
# 优先走 NSSM 服务重启(主程序假死时 HTTP 不可达也能重启成功);
# 未配置服务名时回退环回 HTTP 触发主程序优雅重启
-141
View File
@@ -1,141 +0,0 @@
# -*- coding: utf-8 -*-
"""日志实时推送管理器。
B 端通过 LOG_TAIL_CTRL 命令开启/停止某设备某日志文件的实时流:
- 开启后每 2 秒从主程序 /api/admin/log/tail 拉取增量行
- 新行打包为 LOG_BATCH 上报 B 端(B 端转发给浏览器展示)
- 每个文件独立游标(已发送行数),支持多文件并行
增量算法:tail(max(offset+FETCH, 400)) 行,取 offset 之后的部分;
日志滚动(文件变小/重置)时自动重置游标。
"""
import json
import logging
import threading
import time
log = logging.getLogger(__name__)
# 轮询间隔(秒)
POLL_SEC = 2
# 单次拉取上限(行)
FETCH_MAX = 400
class LogStreamManager:
"""日志实时流:单线程轮询全部活跃文件,增量上报 LOG_BATCH。"""
def __init__(self, config, main_app_client, client):
self.config = config
self.main_app_client = main_app_client
self.client = client
self._cursors: dict[str, int] = {} # 文件名 → 已发送行数
self._active: dict[str, bool] = {} # 文件名 → 是否激活
self._lock = threading.Lock()
self._stop = threading.Event()
self._thread: threading.Thread | None = None
# ================= 控制入口 =================
def control(self, action: str, file: str) -> dict:
"""
控制命令入口。
:param action: start / stop
:param file: 日志文件名(info.log / error.log
"""
if action == "start":
with self._lock:
if self._active.get(file):
return {"ok": True, "msg": f"流已存在:{file}"}
self._active[file] = True
self._cursors.pop(file, None)
self._ensure_thread()
log.info("日志流开启:%s", file)
return {"ok": True, "msg": f"日志流已开启:{file}"}
if action == "stop":
with self._lock:
self._active.pop(file, None)
self._cursors.pop(file, None)
log.info("日志流停止:%s", file)
return {"ok": True, "msg": f"日志流已停止:{file}"}
return {"ok": False, "msg": f"未知 action{action}(支持 start/stop"}
def stop_all(self):
"""停止全部流(Agent 停机时调用)。"""
self._stop.set()
with self._lock:
self._active.clear()
self._cursors.clear()
def _ensure_thread(self):
"""确保轮询线程在运行(首个流开启时启动)。"""
if self._thread is not None and self._thread.is_alive():
return
self._thread = threading.Thread(target=self._poll_loop, name="log-stream", daemon=True)
self._thread.start()
# ================= 轮询 =================
def _poll_loop(self):
"""轮询主循环:每 POLL_SEC 秒拉一轮全部活跃文件。"""
while not self._stop.wait(POLL_SEC):
with self._lock:
files = list(self._active.keys())
for file in files:
if self._stop.is_set():
return
try:
self._poll_one(file)
except Exception as e:
log.warning("日志流轮询失败(%s):%s", file, e)
def _poll_one(self, file: str):
"""单轮增量拉取:取尾部行,比对游标,新行打包上报。"""
with self._lock:
if not self._active.get(file):
return
offset = self._cursors.get(file, 0)
# 多拉一段,保证能覆盖到 offset 之前的已有内容
fetch = max(offset + 100, FETCH_MAX)
raw = self.main_app_client.log_tail(file, fetch)
log.debug("日志流轮询:file=%s, offset=%s, fetch=%s, rawLen=%s",
file, offset, fetch, len(raw) if raw else -1)
if raw is None:
return
try:
node = json.loads(raw)
except Exception:
return
if node.get("code") != 0:
return
content = node.get("data")
if not content:
return
all_lines = content.split("\n")
# 日志滚动检测:行数比游标还少很多 → 文件已重置
if len(all_lines) < offset:
with self._lock:
self._cursors[file] = 0
return
if len(all_lines) <= offset:
return # 无新行
fresh = [line for line in all_lines[offset:] if line and line.strip()]
with self._lock:
self._cursors[file] = len(all_lines)
if not fresh:
return
self.client.send("LOG_BATCH", {
"file": file,
"lines": fresh,
"ts": int(time.time()),
})
+1 -3
View File
@@ -89,10 +89,8 @@ def _setup_logging(log_dir: str):
def _shutdown(signum, frame):
"""优雅停机:释放日志流 → 停守护 → 断连接 → 退出。"""
"""优雅停机:停守护 → 断连接 → 退出。"""
log.info("Agent 停机中...signal=%s", signum)
if _dispatcher is not None:
_dispatcher.shutdown_log_stream()
if _guard is not None:
_guard.stop()
if _client is not None:
+19 -1
View File
@@ -9,6 +9,7 @@
import logging
import subprocess
import threading
import time
log = logging.getLogger(__name__)
@@ -17,6 +18,11 @@ log = logging.getLogger(__name__)
# 避免升级替换 jar 时被看门狗拉起的新进程抢占文件锁。
_upgrading = threading.Event()
# 拉起后的宽限期(秒):主程序重启后 JVM 启动 + DB 建连可能耗时数分钟
# (如机器休眠唤醒后网络未就绪,实测 4 分钟)。宽限期内探测失败只记录不累计,
# 避免重启循环(重启→DB 慢→又判宕机→再重启)。
GRACE_PERIOD_SEC = 300
class ProcessGuard:
"""看门狗:周期健康探测 + 宕机拉起 + 告警上报。"""
@@ -28,6 +34,7 @@ class ProcessGuard:
self._fail_count = 0 # 连续健康探测失败次数
self._restart_count = 0 # 连续拉起次数
self._grace_until = 0.0 # 拉起后的宽限期截止时刻(monotonic 秒)
self._stop = threading.Event()
self._thread: threading.Thread | None = None
@@ -57,10 +64,19 @@ class ProcessGuard:
log.warning("守护探测异常:%s", e)
def _check(self):
"""单次探测:健康则清零计数;连续 3 次失败则拉起(升级期间跳过)。"""
"""单次探测:健康则清零计数;连续 3 次失败则拉起(升级期间/宽限期内跳过)。"""
health = self.main_app_client.health()
if health.get("ok") is True:
self._fail_count = 0
if self._grace_until and time.monotonic() < self._grace_until:
log.info("宽限期内主程序恢复健康,结束宽限")
self._grace_until = 0.0
return
# 拉起后宽限期内:主程序可能在启动/DB 建连中,只记录不累计失败
if self._grace_until and time.monotonic() < self._grace_until:
log.info("主程序拉起后宽限期内(剩 %d 秒),本次探测失败不累计:%s",
int(self._grace_until - time.monotonic()), health.get("error"))
return
# 升级流程主动停的主程序,看门狗不接管,避免抢占 jar 锁
@@ -105,6 +121,8 @@ class ProcessGuard:
else:
proc = start_main_app(self.config.main_start_cmd)
log.info("启动命令已执行,pid=%s", proc.pid if proc else None)
# 拉起后进入宽限期:JVM 启动 + DB 建连期间不累计探测失败,防重启循环
self._grace_until = time.monotonic() + GRACE_PERIOD_SEC
except Exception as e:
log.error("拉起失败:%s", e)
+26
View File
@@ -21,6 +21,7 @@ import shutil
import subprocess
import threading
import time
import zipfile
import requests
@@ -118,6 +119,18 @@ class UpgradeCommand:
return self._fail(version, f"SHA-256 校验失败:expected={sha256}, actual={actual}")
log.info("SHA-256 校验通过")
# ===== 2.5 解压 zip 得到内层 jar(上传端为绕过中间设备拦截,jar 被打成 zip) =====
extracted_jar = os.path.join("tmp", f"upgrade-{version}.jar.extracted")
try:
result = extract_jar_from_zip(tmp_jar, extracted_jar)
if result is None:
os.remove(tmp_jar)
return self._fail(version, "zip 包内未找到 jar 文件")
os.remove(tmp_jar) # 删除临时 zip
tmp_jar = extracted_jar # 后续替换使用解压出的真 jar
except Exception as e:
return self._fail(version, f"解压 zip 失败:{e}")
# ===== 3. 备份 =====
# jar 路径来自配置(main.jar.path),适配不同部署结构
#(如 Agent 与主程序分目录、jar 文件名带版本号等)
@@ -333,6 +346,19 @@ def copy_with_retry(src: str, dst: str, retries: int):
) from last_error
def extract_jar_from_zip(zip_path: str, dst_jar: str) -> str | None:
"""从 zip 中取出第一个 .jar 文件写到 dst_jar,返回 dst_jar;找不到返回 None。"""
with zipfile.ZipFile(zip_path) as z:
jars = [n for n in z.namelist()
if n.lower().endswith(".jar") and not n.endswith("/")]
if not jars:
return None
with z.open(jars[0]) as src, open(dst_jar, "wb") as dst:
shutil.copyfileobj(src, dst)
log.info("解压 zip 得到 jar%s%s", jars[0], os.path.basename(dst_jar))
return dst_jar
def sha256_file(path: str) -> str:
"""文件 SHA-256(分块读取,支持大文件)。"""
digest = hashlib.sha256()