Add graceful SCUM stop, restart, and version update flow
SCUM 停止/重启/更新以前只有“结束进程”这一条路,插件没有声明任何优雅关闭方式,
平台也没有把停止后重新启动串起来。现在插件声明自己的关闭脚本,run 先执行它,
平台在停止或更新成功后再自动拉起服务。
run:
- lifecycle stop 支持插件声明的 gracefulStop(可执行文件、参数、环境、超时、
fallback=report|terminate);关闭命令超时且声明 report 时任务失败,不再默默杀进程。
- 新增 steam.update 依赖探针:调用 steamcmd +app_info_print 获取公开分支 buildid,
与本地 steamapps/appmanifest_<appid>.acf 的 buildid 比较,输出
installed/latest/update=yes|no|unknown。
platform:
- 新增 POST /api/v1/server-instances/{id}/restart 与 /update。
- restart 派发插件 stop 动作(走优雅关闭),终态成功后入队 start 作业。
- update 派发插件 install 动作;插件在更新前必须先优雅关闭 SCUM,关闭失败直接拒绝
SteamCMD 更新,成功后平台再拉起服务。
- 依赖检查输入带上插件声明的服务器安装根目录,供 steam.update 读取 appmanifest。
plugin (SCUM server plugin 0.1.16):
- bin/scum-stop.cmd:解析已声明的可执行文件路径,定位同路径正在运行的 SCUMServer.exe,
通过本地 RCON 公告并发送关闭命令,等待进程自行退出;不再使用 taskkill。
- bin/scum-rcon.ps1:插件自有的 Source RCON 客户端,从 UE4SS mod config.ini 读取
密码/端口,密钥不离开本机。
- actions/stop.json 声明 gracefulStop;actions/install.json 更新前先执行同一关闭脚本。
platform_web:
- 服务器详情新增“重启”按钮和“SCUM 版本更新”面板;点“检查更新”查询公开分支版本,
只有检测到更新时“更新版本”按钮才会置为可用并高亮,点击后先确认再派发更新任务。
This commit is contained in:
@@ -1,6 +1,7 @@
|
||||
package service
|
||||
|
||||
import (
|
||||
"log"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
@@ -113,9 +114,39 @@ func (svc *CoreService) projectLifecycleJobResult(job domain.Job, stamp time.Tim
|
||||
return err
|
||||
}
|
||||
svc.publishLogProcessState(instance)
|
||||
svc.dispatchFollowUpLifecycle(job, instance)
|
||||
return nil
|
||||
}
|
||||
|
||||
// dispatchFollowUpLifecycle starts the server again after a declared graceful
|
||||
// stop or a completed game update. A failed or cancelled stop/update leaves the
|
||||
// instance in its projected state so the operator sees the failure before any
|
||||
// start is attempted.
|
||||
func (svc *CoreService) dispatchFollowUpLifecycle(job domain.Job, instance domain.ServerInstance) {
|
||||
if job.State != domain.JobStateSucceeded {
|
||||
return
|
||||
}
|
||||
operation := job.ExecutionInput.LifecycleOperation
|
||||
if operation != string(domain.ServerLifecycleActionRestart) && operation != string(domain.ServerLifecycleActionUpdate) {
|
||||
return
|
||||
}
|
||||
command := domain.ServerLifecycleCommand{
|
||||
ServerInstanceID: instance.ID,
|
||||
ExpectedConfigVersion: instance.ConfigVersion,
|
||||
IdempotencyKey: job.IdempotencyKey + ":start",
|
||||
}
|
||||
result, err := svc.dispatchExistingServerLifecycle(command, domain.ServerLifecycleActionStart, []domain.ServerInstanceState{
|
||||
domain.ServerInstanceStateReady,
|
||||
domain.ServerInstanceStateStopped,
|
||||
domain.ServerInstanceStateFailed,
|
||||
})
|
||||
if err != nil {
|
||||
log.Printf("PLATFORM phase=lifecycle status=follow_up_failed operation=%s server=%s error=%s", safeLogValue(operation), safeLogValue(instance.ID), safeLogValue(err.Error()))
|
||||
return
|
||||
}
|
||||
log.Printf("PLATFORM phase=lifecycle status=follow_up_queued operation=%s server=%s job=%s", safeLogValue(operation), safeLogValue(instance.ID), safeLogValue(result.Job.ID))
|
||||
}
|
||||
|
||||
func (svc *CoreService) projectServerDeploymentProgress(job domain.Job, stamp time.Time) error {
|
||||
if job.ExecutionInput.Deployment == nil || job.ServerInstanceID == "" {
|
||||
return nil
|
||||
|
||||
Reference in New Issue
Block a user