(dubbo-kubernetes) branch master updated: [horus] Execute a scripted reboot operation (#412)

[zhongxjian]

This is an automated email from the ASF dual-hosted git repository.

zhongxjian pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/dubbo-kubernetes.git


The following commit(s) were added to refs/heads/master by this push:
     new 95d9f141 [horus] Execute a scripted reboot operation (#412)
95d9f141 is described below

commit 95d9f141683d1fb16138ec791bf7b890d5657035
Author: mfordjody <11638...@qq.com>
AuthorDate: Tue Oct 1 15:09:09 2024 +0800

    [horus] Execute a scripted reboot operation (#412)
---
 app/horus/core/horuser/node_downtime.go |  1 -
 app/horus/core/horuser/node_restart.go  | 14 ++++++++++++--
 2 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/app/horus/core/horuser/node_downtime.go 
b/app/horus/core/horuser/node_downtime.go
index 3c028786..eea3bc2d 100644
--- a/app/horus/core/horuser/node_downtime.go
+++ b/app/horus/core/horuser/node_downtime.go
@@ -134,7 +134,6 @@ func (h *Horuser) DownTimeNodes(clusterName, addr string) {
                nodeIP, err := func() (string, error) {
                        for _, address := range node.Status.Addresses {
                                if address.Type == "InternalIP" {
-
                                        return address.Address, nil
                                }
                        }
diff --git a/app/horus/core/horuser/node_restart.go 
b/app/horus/core/horuser/node_restart.go
index b398eca6..743ea7c0 100644
--- a/app/horus/core/horuser/node_restart.go
+++ b/app/horus/core/horuser/node_restart.go
@@ -23,6 +23,7 @@ import (
        "github.com/gammazero/workerpool"
        "k8s.io/apimachinery/pkg/util/wait"
        "k8s.io/klog/v2"
+       "os/exec"
        "time"
 )
 
@@ -64,8 +65,17 @@ func (h *Horuser) TryRestart(node db.NodeDataInfo) {
        klog.Infof("RestartMarker result pass:%v err:%v", pass, err)
 
        if pass {
-               msg := fmt.Sprintf("【等待腾空节点后重启就绪】【节点:%v】【日期:%v】【集群:%v】", 
node.NodeName, node.FirstDate, node.ClusterName)
+               msg := fmt.Sprintf("【等待宕机节点腾空后重启】【节点:%v】【日期:%v】【集群:%v】", 
node.NodeName, node.FirstDate, node.ClusterName)
                alert.DingTalkSend(h.cc.NodeDownTime.DingTalk, msg)
-               // TODO restart policy
+               // TODO user@password
+               cmd := exec.Command("/bin/bash", "./restart.sh", node.NodeIP)
+               output, err := cmd.CombinedOutput()
+               if err != nil {
+                       klog.Errorf("Failed to execute restart.sh script for 
node %v: %v", node.NodeName, err)
+                       klog.Errorf("Output: %v", string(output))
+                       return
+               }
+               klog.Infof("Successfully executed restart.sh for node %v. 
Output: %v", node.NodeName, string(output))
+       }
        }
 }