fix(ci): vps_restart_resident schtasks /run 加3次重试——run756实锤三连部署短时海量ssh连接可被sshd单发重置(kex reset),拉起命令没送到=影子舰队整夜静默需手动补拉 [nas]

2026-08-26 22:49:06 +08:00
parent c16f5e2248
commit a7c2dd6f59
1 changed files with 14 additions and 2 deletions
@@ -36,8 +36,20 @@ $SSH "schtasks /end /tn sanguo-shadow-desk" 2>/dev/null || true

 # ---- 4) 重拉 ----
 sleep 3
-$SSH "schtasks /run /tn sanguo-live-supervisor" || { echo "❌ supervisor 拉起失败"; exit 1; }
-$SSH "schtasks /run /tn sanguo-shadow-desk" || { echo "❌ shadow-desk 拉起失败"; exit 1; }
+# schtasks /run 带重试(2026-08-26 run756 实锤: 短时海量 ssh 连接可被 sshd 重置
+# `kex_exchange_identification: Connection reset`——单发失败=影子舰队整夜静默,
+# 需手动 schtasks /run 补拉; 3 次重试吃掉瞬时抖动)
+run_task() {
+  local tn=$1 i
+  for i in 1 2 3; do
+    $SSH "schtasks /run /tn $tn" && return 0
+    echo "  (拉起 $tn 第${i}次失败, 5s 后重试)"
+    sleep 5
+  done
+  return 1
+}
+run_task sanguo-live-supervisor || { echo "❌ supervisor 拉起失败"; exit 1; }
+run_task sanguo-shadow-desk || { echo "❌ shadow-desk 拉起失败"; exit 1; }

 # ---- 5) 等进程回来（内存实测 40-60s 起齐，给 120s 余量）----
 for i in $(seq 1 12); do