feat: 增加升级失败回滚操作

.idea/.gitignore

+# Default ignored files
+/shelf/
+/workspace.xml
+# Editor-based HTTP Client requests
+/httpRequests/
+# Datasource local storage ignored files
+/dataSources/
+/dataSources.local.xml

.idea/modules.xml

+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectModuleManager">
+    <modules>
+      <module fileurl="file://$PROJECT_DIR$/.idea/public.iml" filepath="$PROJECT_DIR$/.idea/public.iml" />
+    </modules>
+  </component>
+</project>
\ No newline at end of file

.idea/public.iml

+<?xml version="1.0" encoding="UTF-8"?>
+<module type="WEB_MODULE" version="4">
+  <component name="Go" enabled="true" />
+  <component name="NewModuleRootManager">
+    <content url="file://$MODULE_DIR$" />
+    <orderEntry type="inheritedJdk" />
+    <orderEntry type="sourceFolder" forTests="false" />
+  </component>
+</module>
\ No newline at end of file

.idea/vcs.xml

+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="VcsDirectoryMappings">
+    <mapping directory="" vcs="Git" />
+  </component>
+</project>
\ No newline at end of file

lib/monitor_daemon/monitor_daemon.go

@@ -14,6 +14,7 @@ import (
 	"time"
 
 	"linkfog.com/public/lib/common"
+	"linkfog.com/public/lib/file"
 	"linkfog.com/public/lib/l"
 )
 
@@ -43,6 +44,8 @@ type MonitorDaemon struct {
 	restartInterval *restartInterval
 	cmdline         string
 	childPid        int
+	agentMd5        string
+	cmdline0        string
 
 	isProcAbnormalCallback IsProcAbnormalCallback
 }
@@ -146,8 +149,16 @@ func New(cmdline string, opts ...MonitorDaemonOpt) (*MonitorDaemon, error) {
 		opt(&m)
 	}
 
+	m.cmdline0 = strings.Split(m.cmdline, " ")[0]
+	md5, err := file.GetFileMD5(m.cmdline0)
+	if err != nil {
+		l.Errorf("calculation of agent MD5(%s) failed err:%s", m.cmdline0, err)
+	} else {
+		m.agentMd5 = md5
+	}
+
 	//第一次启动进程
-	err := m.startProcess()
+	err = m.startProcess()
 	if err != nil {
 		l.Errorf("start child process failed %s err:%v", cmdline, err)
 		return nil, err
@@ -243,29 +254,32 @@ func (m *MonitorDaemon) runWithSignal(sigs []os.Signal) {
 			}
 
 			if !isAlive && m.restartInterval.isNeedStart() {
-				//忽略错误，只打印日志
-				err := m.startProcess()
+
+				md5, err := file.GetFileMD5(m.cmdline0)
 				if err != nil {
+					l.Errorf("calculation of agent MD5(%s) failed err:%s", m.cmdline0, err)
+				}
+				err = m.startProcess()
+				if err != nil {
+					// 异常情况，直接无法启动，进行回滚操作
 					errRetryTimes++
 					l.Errorf("monitorProcess start child process failed %s, errRetryTimes:%d, err:%v", m.cmdline, errRetryTimes, err)
-					if errRetryTimes > m.maxRetryTimes {
-						if _, err := os.Stat(m.agentBackupPath); err == nil || !os.IsNotExist(err) {
-							agentName := strings.Split(m.cmdline, " ")[0]
-							if agentName != "" {
-								err := os.Rename(m.agentBackupPath, agentName)
-								if err != nil {
-									l.Errorf("monitorProcess rollback agent process failed err:%v", err)
-								} else {
-									errRetryTimes = 0
-								}
-							}
-						} else {
+					if md5 != m.agentMd5 && errRetryTimes > m.maxRetryTimes {
+						m.rollbackForAbnormalCase()
+						errRetryTimes = 0
+					}
+				} else {
+					if md5 == m.agentMd5 {
+						errRetryTimes++
+						// 异常情况，启动几秒后停止，进行回滚操作
+						if errRetryTimes > m.maxRetryTimes {
+							m.rollbackForAbnormalCase()
 							errRetryTimes = 0
 						}
+					} else {
+						m.agentMd5 = md5
 					}
-				} else {
-					errRetryTimes = 0
-					l.Infof("monitorProcess start child process success pid: %d", m.childPid)
+					l.Infof("monitorProcess start child process success pid: %d, errRetryTimes:%d", m.childPid, errRetryTimes)
 				}
 				m.restartInterval.setLastStart()
 			}
@@ -435,3 +449,21 @@ func readStderr(stderrReader io.Reader) {
 	}
 }
 
+func (m *MonitorDaemon) rollbackForAbnormalCase() {
+	if _, err := os.Stat(m.agentBackupPath); err == nil || !os.IsNotExist(err) {
+		if m.cmdline0 != "" {
+			err := os.Rename(m.agentBackupPath, m.cmdline0)
+			if err != nil {
+				l.Errorf("monitorProcess rollback agent process failed err:%v", err)
+			} else {
+				md5, err := file.GetFileMD5(m.cmdline0)
+				if err != nil {
+					l.Errorf("calculation of agent MD5(%s) failed err:%s", m.cmdline0, err)
+				} else {
+					m.agentMd5 = md5
+				}
+				l.Infof("monitorProcess rollback agent process successful md5:%s", md5)
+			}
+		}
+	}
+}