fix

terraform/tfs/workers-bak.tf

@@ -0,0 +1,101 @@
+resource "null_resource" "k3s_cleanup_worker1" {
+  # SSH 登录到 worker 节点
+  connection {
+    type        = "ssh"
+    host        = var.worker_ips[0]
+    user        = "root"
+    password    = var.worker_password
+  }
+
+  provisioner "remote-exec" {
+    inline = [
+      "# Stop k3s-agent service if running",
+      "systemctl stop k3s-agent 2>/dev/null || true",
+      "# Kill any remaining k3s processes",
+      "pkill -f k3s 2>/dev/null || true",
+      "# Remove k3s data directory",
+      "rm -rf /var/lib/rancher/k3s",
+      "# Remove k3s-agent service file",
+      "rm -f /etc/systemd/system/k3s-agent.service",
+      "echo 'Cleanup completed on worker node 1'"
+    ]
+  }
+}
+
+resource "null_resource" "k3s_cleanup_worker2" {
+  connection {
+    type        = "ssh"
+    host        = var.worker_ips[1]
+    user        = "root"
+    password    = var.worker_password
+  }
+
+  provisioner "remote-exec" {
+    inline = [
+      "# Stop k3s-agent service if running",
+      "systemctl stop k3s-agent 2>/dev/null || true",
+      "# Kill any remaining k3s processes",
+      "pkill -f k3s 2>/dev/null || true",
+      "# Remove k3s data directory",
+      "rm -rf /var/lib/rancher/k3s",
+      "# Remove k3s-agent service file",
+      "rm -f /etc/systemd/system/k3s-agent.service",
+      "echo 'Cleanup completed on worker node 2'"
+    ]
+  }
+}
+
+resource "null_resource" "k3s_install_worker1" {
+  depends_on = [null_resource.copy_token_to_workers, null_resource.k3s_cleanup_worker1]
+
+  connection {
+    type        = "ssh"
+    host        = var.worker_ips[0]
+    user        = "root"
+    password    = var.worker_password
+  }
+
+  provisioner "remote-exec" {
+    inline = [
+      "if [ -f /usr/local/bin/k3s ]; then",
+      "  echo 'k3s binary already exists, skipping download'",
+      "else",
+      "  wget -O /usr/local/bin/k3s ${var.k3s_download_url}",
+      "  chmod +x /usr/local/bin/k3s",
+      "fi",
+      "TOKEN=$(cat /root/node-token)",
+      "echo -e '\\033[32m--Start k3s agent 1--\\033[0m'",
+      "k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[0], ".", "-")} --node-external-ip=${var.worker_ips[0]} --data-dir /var/lib/rancher/k3s",
+      "echo -e '\\033[32m--Done k3s agent 1--\\033[0m'",
+      "sleep 5",
+    ]
+  }
+}
+
+
+resource "null_resource" "k3s_install_worker2" {
+  depends_on = [null_resource.copy_token_to_workers, null_resource.k3s_cleanup_worker2]
+
+  connection {
+    type        = "ssh"
+    host        = var.worker_ips[1]
+    user        = "root"
+    password    = var.worker_password
+  }
+
+  provisioner "remote-exec" {
+    inline = [
+      "if [ -f /usr/local/bin/k3s ]; then",
+      "  echo 'k3s binary already exists, skipping download'",
+      "else",
+      "  wget -O /usr/local/bin/k3s ${var.k3s_download_url}",
+      "  chmod +x /usr/local/bin/k3s",
+      "fi",
+      "TOKEN=$(cat /root/node-token)",
+      "echo -e '\\033[32m--Start k3s agent 2--\\033[0m'",
+      "k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[1], ".", "-")} --node-external-ip=${var.worker_ips[1]} --data-dir /var/lib/rancher/k3s",
+      "echo -e '\\033[32m--Done k3s agent 2--\\033[0m'",
+      "sleep 5",
+    ]
+  }
+}

terraform/tfs/workers.tf

@@ -64,10 +64,25 @@ resource "null_resource" "k3s_install_worker1" {
       "  chmod +x /usr/local/bin/k3s",
       "fi",
       "TOKEN=$(cat /root/node-token)",
-      "echo -e '\\033[32m--Start k3s agent 1--\\033[0m'",
-      "k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[0], ".", "-")} --node-external-ip=${var.worker_ips[0]} --data-dir /var/lib/rancher/k3s & disown",
-      "echo -e '\\033[32m--Done k3s agent 1--\\033[0m'",
-      "sleep 5",
+      "echo -e '\\033[32m--Install k3s-agent systemd service--\\033[0m'",
+      # 注册 systemd 服务
+      "cat > /etc/systemd/system/k3s-agent.service <<EOF",
+      "[Unit]",
+      "Description=K3s Agent",
+      "After=network.target",
+      "[Service]",
+      "ExecStart=/usr/local/bin/k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[0], \".\", \"-\")} --node-external-ip=${var.worker_ips[0]} --data-dir /var/lib/rancher/k3s",
+      "Restart=always",
+      "KillMode=process",
+      "Delegate=yes",
+      "LimitNOFILE=65536",
+      "[Install]",
+      "WantedBy=multi-user.target",
+      "EOF",
+      "systemctl daemon-reload",
+      "systemctl enable --now k3s-agent",
+      "echo -e '\\033[32m--k3s agent started successfully--\\033[0m'",
+      "sleep 3",
     ]
   }
 }
@@ -92,10 +107,25 @@ resource "null_resource" "k3s_install_worker2" {
       "  chmod +x /usr/local/bin/k3s",
       "fi",
       "TOKEN=$(cat /root/node-token)",
-      "echo -e '\\033[32m--Start k3s agent 2--\\033[0m'",
-      "k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[1], ".", "-")} --node-external-ip=${var.worker_ips[1]} --data-dir /var/lib/rancher/k3s & disown",
-      "echo -e '\\033[32m--Done k3s agent 2--\\033[0m'",
-      "sleep 5",
+      "echo -e '\\033[32m--Install k3s-agent systemd service--\\033[0m'",
+      # 注册 systemd 服务
+      "cat > /etc/systemd/system/k3s-agent.service <<EOF",
+      "[Unit]",
+      "Description=K3s Agent",
+      "After=network.target",
+      "[Service]",
+      "ExecStart=/usr/local/bin/k3s agent --server https://${var.master_ip}:6443 --token $TOKEN --node-name worker-node-${replace(var.worker_ips[0], \".\", \"-\")} --node-external-ip=${var.worker_ips[1]} --data-dir /var/lib/rancher/k3s",
+      "Restart=always",
+      "KillMode=process",
+      "Delegate=yes",
+      "LimitNOFILE=65536",
+      "[Install]",
+      "WantedBy=multi-user.target",
+      "EOF",
+      "systemctl daemon-reload",
+      "systemctl enable --now k3s-agent",
+      "echo -e '\\033[32m--k3s agent started successfully--\\033[0m'",
+      "sleep 3",
     ]
   }
 }