Files
aiops/backend/internal/api/handlers/automation.go
T

714 lines
18 KiB
Go

package handlers
import (
"bytes"
"context"
"fmt"
"net/http"
"os/exec"
"strconv"
"time"
"ezops/internal/models"
"github.com/gin-gonic/gin"
"gorm.io/gorm"
)
// runLocal executes a script or shell command locally on the AIOps host and
// returns the combined stdout+stderr output, the process exit code (0 = success)
// and the wall-clock duration in milliseconds. A timeout prevents runaway jobs.
func runLocal(scriptType, content string, timeoutSec int) (output string, exitCode int, durationMs int) {
start := time.Now()
timeout := time.Duration(timeoutSec) * time.Second
if timeout <= 0 {
timeout = 300 * time.Second
}
ctx, cancel := context.WithTimeout(context.Background(), timeout)
defer cancel()
var cmd *exec.Cmd
switch scriptType {
case "python":
cmd = exec.CommandContext(ctx, "python3", "-c", content)
default: // shell / command / "" → run via bash
cmd = exec.CommandContext(ctx, "bash", "-c", content)
}
var out bytes.Buffer
cmd.Stdout = &out
cmd.Stderr = &out
err := cmd.Run()
durationMs = int(time.Since(start).Milliseconds())
output = out.String()
if ctx.Err() == context.DeadlineExceeded {
output += fmt.Sprintf("\n[超时] 执行超过 %d 秒被终止", timeoutSec)
return output, -1, durationMs
}
if err != nil {
if exitError, ok := err.(*exec.ExitError); ok {
exitCode = exitError.ExitCode()
} else {
exitCode = -1
output += "\n" + err.Error()
}
return output, exitCode, durationMs
}
return output, 0, durationMs
}
// AutomationHandler 自动化运维处理器
type AutomationHandler struct {
db *gorm.DB
}
// NewAutomationHandler 创建自动化运维处理器
func NewAutomationHandler(db *gorm.DB) *AutomationHandler {
return &AutomationHandler{db: db}
}
// ListCronJobs 获取定时任务列表
// @Summary 获取定时任务列表
// @Description 获取所有定时任务列表
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param page query int false "页码"
// @Param page_size query int false "每页数量"
// @Success 200 {object} models.Response{data=models.ListResponse}
// @Router /cron-jobs [get]
func (h *AutomationHandler) ListCronJobs(c *gin.Context) {
var query models.PageQuery
c.ShouldBindQuery(&query)
var total int64
h.db.Model(&models.CronJob{}).Count(&total)
var jobs []models.CronJob
h.db.Preload("Script").Offset(query.GetOffset()).Limit(query.GetLimit()).Find(&jobs)
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "success",
Data: models.ListResponse{
List: jobs,
Total: total,
},
})
}
// GetCronJob 获取定时任务详情
// @Summary 获取定时任务详情
// @Description 获取指定定时任务的详细信息
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "任务ID"
// @Success 200 {object} models.Response{data=models.CronJob}
// @Router /cron-jobs/{id} [get]
func (h *AutomationHandler) GetCronJob(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var job models.CronJob
if err := h.db.Preload("Script").First(&job, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "定时任务不存在",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "success",
Data: job,
})
}
// CreateCronJob 创建定时任务
// @Summary 创建定时任务
// @Description 创建新定时任务
// @Tags 自动化运维
// @Accept json
// @Produce json
// @Security Bearer
// @Param request body models.CreateCronJobRequest true "创建请求"
// @Success 200 {object} models.Response{data=models.CronJob}
// @Router /cron-jobs [post]
func (h *AutomationHandler) CreateCronJob(c *gin.Context) {
var req models.CreateCronJobRequest
if err := c.ShouldBindJSON(&req); err != nil {
c.JSON(http.StatusBadRequest, models.Response{
Code: 400,
Message: "请求参数错误: " + err.Error(),
})
return
}
job := models.CronJob{
Name: req.Name,
Description: req.Description,
CronExpr: req.CronExpr,
Type: req.Type,
Content: req.Content,
ScriptID: req.ScriptID,
TargetType: req.TargetType,
Timeout: req.Timeout,
NotifyOn: req.NotifyOn,
Enabled: true,
}
if job.Timeout == 0 {
job.Timeout = 300
}
if job.NotifyOn == "" {
job.NotifyOn = "failure"
}
if err := h.db.Create(&job).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "创建失败",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "创建成功",
Data: job,
})
}
// UpdateCronJob 更新定时任务
// @Summary 更新定时任务
// @Description 更新指定定时任务
// @Tags 自动化运维
// @Accept json
// @Produce json
// @Security Bearer
// @Param id path int true "任务ID"
// @Param request body models.CreateCronJobRequest true "更新请求"
// @Success 200 {object} models.Response{data=models.CronJob}
// @Router /cron-jobs/{id} [put]
func (h *AutomationHandler) UpdateCronJob(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var req models.CreateCronJobRequest
if err := c.ShouldBindJSON(&req); err != nil {
c.JSON(http.StatusBadRequest, models.Response{
Code: 400,
Message: "请求参数错误: " + err.Error(),
})
return
}
var job models.CronJob
if err := h.db.First(&job, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "定时任务不存在",
})
return
}
updates := map[string]interface{}{
"name": req.Name,
"description": req.Description,
"cron_expr": req.CronExpr,
"type": req.Type,
"content": req.Content,
"script_id": req.ScriptID,
"target_type": req.TargetType,
"timeout": req.Timeout,
"notify_on": req.NotifyOn,
}
if err := h.db.Model(&job).Updates(updates).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "更新失败",
})
return
}
h.db.Preload("Script").First(&job, id)
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "更新成功",
Data: job,
})
}
// DeleteCronJob 删除定时任务
// @Summary 删除定时任务
// @Description 删除指定定时任务
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "任务ID"
// @Success 200 {object} models.Response
// @Router /cron-jobs/{id} [delete]
func (h *AutomationHandler) DeleteCronJob(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
if err := h.db.Delete(&models.CronJob{}, id).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "删除失败",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "删除成功",
})
}
// ToggleCronJob 切换定时任务状态
// @Summary 切换定时任务状态
// @Description 启用或禁用定时任务
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "任务ID"
// @Success 200 {object} models.Response{data=models.CronJob}
// @Router /cron-jobs/{id}/toggle [post]
func (h *AutomationHandler) ToggleCronJob(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var job models.CronJob
if err := h.db.First(&job, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "定时任务不存在",
})
return
}
job.Enabled = !job.Enabled
if err := h.db.Save(&job).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "操作失败",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "操作成功",
Data: job,
})
}
// ExecuteCronJob 执行定时任务
// @Summary 执行定时任务
// @Description 手动执行定时任务
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "任务ID"
// @Success 200 {object} models.Response{data=models.TaskExecution}
// @Router /cron-jobs/{id}/execute [post]
func (h *AutomationHandler) ExecuteCronJob(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var job models.CronJob
if err := h.db.First(&job, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "定时任务不存在",
})
return
}
// Resolve the content to execute. For type=script, load the linked Script.
content := job.Content
scriptType := "shell"
if job.Type == "script" && job.ScriptID != nil {
var script models.Script
if err := h.db.First(&script, *job.ScriptID).Error; err == nil {
content = script.Content
scriptType = script.Type
h.db.Model(&script).UpdateColumn("run_count", script.RunCount+1)
}
}
now := time.Now()
execution := models.TaskExecution{
TaskType: "cron",
TaskID: uint(id),
TaskName: job.Name,
Status: "running",
TargetType: job.TargetType,
Targets: job.Targets,
StartedAt: &now,
}
if err := h.db.Create(&execution).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "创建执行记录失败",
})
return
}
// Execute locally on the AIOps host.
output, exitCode, duration := runLocal(scriptType, content, job.Timeout)
finishedAt := time.Now()
status := "success"
if exitCode != 0 {
status = "failed"
}
h.db.Model(&execution).Updates(map[string]interface{}{
"status": status,
"output": output,
"duration": duration,
"finished_at": finishedAt,
})
// Update cron job statistics.
statsUpdates := map[string]interface{}{
"last_run_at": now,
"last_status": status,
"run_count": job.RunCount + 1,
}
if status == "success" {
statsUpdates["success_count"] = job.SuccessCount + 1
} else {
statsUpdates["fail_count"] = job.FailCount + 1
}
h.db.Model(&job).Updates(statsUpdates)
execution.Output = output
execution.Status = status
execution.Duration = duration
execution.FinishedAt = &finishedAt
msg := "执行成功"
if status == "failed" {
msg = "执行失败"
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: msg,
Data: execution,
})
}
// ListScripts 获取脚本列表
// @Summary 获取脚本列表
// @Description 获取所有脚本列表
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param page query int false "页码"
// @Param page_size query int false "每页数量"
// @Success 200 {object} models.Response{data=models.ListResponse}
// @Router /scripts [get]
func (h *AutomationHandler) ListScripts(c *gin.Context) {
var query models.PageQuery
c.ShouldBindQuery(&query)
var total int64
h.db.Model(&models.Script{}).Count(&total)
var scripts []models.Script
h.db.Offset(query.GetOffset()).Limit(query.GetLimit()).Find(&scripts)
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "success",
Data: models.ListResponse{
List: scripts,
Total: total,
},
})
}
// GetScript 获取脚本详情
// @Summary 获取脚本详情
// @Description 获取指定脚本的详细信息
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "脚本ID"
// @Success 200 {object} models.Response{data=models.Script}
// @Router /scripts/{id} [get]
func (h *AutomationHandler) GetScript(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var script models.Script
if err := h.db.First(&script, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "脚本不存在",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "success",
Data: script,
})
}
// CreateScript 创建脚本
// @Summary 创建脚本
// @Description 创建新脚本
// @Tags 自动化运维
// @Accept json
// @Produce json
// @Security Bearer
// @Param request body models.CreateScriptRequest true "创建请求"
// @Success 200 {object} models.Response{data=models.Script}
// @Router /scripts [post]
func (h *AutomationHandler) CreateScript(c *gin.Context) {
var req models.CreateScriptRequest
if err := c.ShouldBindJSON(&req); err != nil {
c.JSON(http.StatusBadRequest, models.Response{
Code: 400,
Message: "请求参数错误: " + err.Error(),
})
return
}
script := models.Script{
Name: req.Name,
Code: req.Code,
Description: req.Description,
Type: req.Type,
Content: req.Content,
Params: req.Params,
Timeout: req.Timeout,
}
if script.Type == "" {
script.Type = "shell"
}
if script.Timeout == 0 {
script.Timeout = 300
}
if err := h.db.Create(&script).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "创建失败",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "创建成功",
Data: script,
})
}
// UpdateScript 更新脚本
// @Summary 更新脚本
// @Description 更新指定脚本
// @Tags 自动化运维
// @Accept json
// @Produce json
// @Security Bearer
// @Param id path int true "脚本ID"
// @Param request body models.CreateScriptRequest true "更新请求"
// @Success 200 {object} models.Response{data=models.Script}
// @Router /scripts/{id} [put]
func (h *AutomationHandler) UpdateScript(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var req models.CreateScriptRequest
if err := c.ShouldBindJSON(&req); err != nil {
c.JSON(http.StatusBadRequest, models.Response{
Code: 400,
Message: "请求参数错误: " + err.Error(),
})
return
}
var script models.Script
if err := h.db.First(&script, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "脚本不存在",
})
return
}
updates := map[string]interface{}{
"name": req.Name,
"description": req.Description,
"type": req.Type,
"content": req.Content,
"params": req.Params,
"timeout": req.Timeout,
"version": script.Version + 1,
}
if err := h.db.Model(&script).Updates(updates).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "更新失败",
})
return
}
h.db.First(&script, id)
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "更新成功",
Data: script,
})
}
// DeleteScript 删除脚本
// @Summary 删除脚本
// @Description 删除指定脚本
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param id path int true "脚本ID"
// @Success 200 {object} models.Response
// @Router /scripts/{id} [delete]
func (h *AutomationHandler) DeleteScript(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
if err := h.db.Delete(&models.Script{}, id).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "删除失败",
})
return
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "删除成功",
})
}
// ExecuteScript 执行脚本
// @Summary 执行脚本
// @Description 在目标服务器上执行脚本
// @Tags 自动化运维
// @Accept json
// @Produce json
// @Security Bearer
// @Param id path int true "脚本ID"
// @Param request body models.ExecuteScriptRequest true "执行请求"
// @Success 200 {object} models.Response{data=models.TaskExecution}
// @Router /scripts/{id}/execute [post]
func (h *AutomationHandler) ExecuteScript(c *gin.Context) {
id, _ := strconv.Atoi(c.Param("id"))
var script models.Script
if err := h.db.First(&script, id).Error; err != nil {
c.JSON(http.StatusNotFound, models.Response{
Code: 404,
Message: "脚本不存在",
})
return
}
// The request body is optional — when omitted we execute locally.
var req models.ExecuteScriptRequest
_ = c.ShouldBindJSON(&req)
now := time.Now()
execution := models.TaskExecution{
TaskType: "script",
TaskID: uint(id),
TaskName: script.Name,
Status: "running",
TargetType: req.TargetType,
StartedAt: &now,
}
if err := h.db.Create(&execution).Error; err != nil {
c.JSON(http.StatusInternalServerError, models.Response{
Code: 500,
Message: "创建执行记录失败",
})
return
}
// Execute locally on the AIOps host.
output, exitCode, duration := runLocal(script.Type, script.Content, script.Timeout)
finishedAt := time.Now()
status := "success"
if exitCode != 0 {
status = "failed"
}
h.db.Model(&execution).Updates(map[string]interface{}{
"status": status,
"output": output,
"duration": duration,
"finished_at": finishedAt,
})
h.db.Model(&script).UpdateColumn("run_count", script.RunCount+1)
execution.Output = output
execution.Status = status
execution.Duration = duration
execution.FinishedAt = &finishedAt
msg := "执行成功"
if status == "failed" {
msg = "执行失败"
}
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: msg,
Data: execution,
})
}
// ListExecutions 获取任务执行记录列表
// @Summary 获取任务执行记录
// @Description 获取脚本/定时任务的执行历史记录(支持 task_type / task_id 筛选)
// @Tags 自动化运维
// @Produce json
// @Security Bearer
// @Param page query int false "页码"
// @Param page_size query int false "每页数量"
// @Param task_type query string false "任务类型 (cron/script)"
// @Param task_id query string false "任务ID"
// @Success 200 {object} models.Response{data=models.ListResponse}
// @Router /task-executions [get]
func (h *AutomationHandler) ListExecutions(c *gin.Context) {
var query models.PageQuery
c.ShouldBindQuery(&query)
db := h.db.Model(&models.TaskExecution{})
if taskType := c.Query("task_type"); taskType != "" {
db = db.Where("task_type = ?", taskType)
}
if taskID := c.Query("task_id"); taskID != "" {
db = db.Where("task_id = ?", taskID)
}
var total int64
db.Count(&total)
var execs []models.TaskExecution
db.Order("created_at DESC").Offset(query.GetOffset()).Limit(query.GetLimit()).Find(&execs)
c.JSON(http.StatusOK, models.Response{
Code: 200,
Message: "success",
Data: models.ListResponse{
List: execs,
Total: total,
},
})
}