0

doc/server_error_log_monitor.md

@@ -101,6 +101,20 @@ INTERNAL_API_TOKEN=你的token pm2 start scripts/collect-pm2-error-logs.js \
   --update-env
 ```
 
+默认只采集 `app-24-error.log`，避免把旧服务 `app-error.log` 的历史噪音写入新表。若 PM2 进程名变化，可以通过环境变量指定:
+
+```bash
+PM2_ERROR_LOG_NAMES=app-24
+```
+
+可同时指定多个:
+
+```bash
+PM2_ERROR_LOG_NAMES=app-24,another-app
+```
+
+采集脚本也会忽略负载均衡健康检查导致的 `clb-healthcheck + write ECONNRESET` 噪音。
+
 检查状态:
 
 ```bash
@@ -303,6 +317,18 @@ curl -H "x-internal-token: 你的token" \
 
 现在推荐使用 `local-error-monitor/check-once.js` 的 SSH 远程 curl 模式，不需要本机直接访问公网端口。
 
+### 健康检查 ECONNRESET
+
+如果错误类似:
+
+```text
+Error: write ECONNRESET
+user-agent: clb-healthcheck
+GET /
+```
+
+通常是负载均衡健康检查断开连接产生的噪音。当前 `scripts/collect-pm2-error-logs.js` 已默认过滤这类日志。
+
 ### pm2: command not found
 
 非交互 SSH 可能不会加载完整 PATH。排查 PM2 时可以用登录 shell:

scripts/collect-pm2-error-logs.js

@@ -9,6 +9,30 @@ const logDir = process.env.PM2_ERROR_LOG_DIR || path.join(os.homedir(), '.pm2',
 const stateFile = process.env.PM2_ERROR_LOG_STATE_FILE || path.join(process.cwd(), '.pm2-error-log-cursors.json');
 const pollIntervalMs = Number(process.env.PM2_ERROR_LOG_POLL_INTERVAL_MS || 30000);
 const readExisting = process.env.PM2_ERROR_LOG_READ_EXISTING === 'true';
+const includeLogNames = parseNameList(process.env.PM2_ERROR_LOG_NAMES || 'app-24');
+const excludeLogNames = parseNameList(process.env.PM2_ERROR_LOG_EXCLUDE_NAMES || '');
+
+function parseNameList(value) {
+  return new Set(
+    String(value || '')
+      .split(',')
+      .map((item) => item.trim())
+      .filter(Boolean)
+      .flatMap((item) => item.endsWith('-error.log') ? [item] : [item, `${item}-error.log`])
+  );
+}
+
+function shouldCollectLogFile(file) {
+  if (!file.endsWith('-error.log')) {
+    return false;
+  }
+
+  if (excludeLogNames.has(file)) {
+    return false;
+  }
+
+  return includeLogNames.size === 0 || includeLogNames.has(file);
+}
 
 async function loadState() {
   try {
@@ -26,7 +50,7 @@ async function listPm2ErrorLogs() {
   try {
     const files = await fsp.readdir(logDir);
     return files
-      .filter((file) => file.endsWith('-error.log'))
+      .filter(shouldCollectLogFile)
       .map((file) => path.join(logDir, file));
   } catch (error) {
     console.error(`[pm2-error-log-collector] cannot read log dir ${logDir}:`, error.message || error);
@@ -91,6 +115,14 @@ function getMessageFromBlock(block) {
     : 'PM2 error log';
 }
 
+function shouldIgnoreBlock(block) {
+  return (
+    /ECONNRESET/.test(block) &&
+    /write ECONNRESET/.test(block) &&
+    /clb-healthcheck/i.test(block)
+  );
+}
+
 async function collectOnce(state) {
   const files = await listPm2ErrorLogs();
 
@@ -113,6 +145,10 @@ async function collectOnce(state) {
     state[filePath] = stat.size;
 
     for (const block of splitErrorBlocks(content)) {
+      if (shouldIgnoreBlock(block)) {
+        continue;
+      }
+
       await recordError(new Error(getMessageFromBlock(block)), null, {
         source: 'pm2-log',
         stack: block,
@@ -127,6 +163,10 @@ async function collectOnce(state) {
 
 async function main() {
   console.log(`[pm2-error-log-collector] watching ${logDir}`);
+  console.log(`[pm2-error-log-collector] include logs: ${includeLogNames.size ? [...includeLogNames].join(', ') : 'all'}`);
+  if (excludeLogNames.size > 0) {
+    console.log(`[pm2-error-log-collector] exclude logs: ${[...excludeLogNames].join(', ')}`);
+  }
   const state = await loadState();
 
   await collectOnce(state);