Skip to content
5 changes: 4 additions & 1 deletion agent/app/repo/task.go
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@ package repo

import (
"context"
"time"

"github.com/1Panel-dev/1Panel/agent/constant"
"github.com/1Panel-dev/1Panel/agent/global"
Expand Down Expand Up @@ -110,7 +111,9 @@ func (t TaskRepo) Update(ctx context.Context, task *model.Task) error {
}

func (t TaskRepo) UpdateRunningTaskToFailed() error {
return getTaskDb(t.WithByStatus(constant.StatusExecuting)).Model(&model.Task{}).Updates(map[string]interface{}{"status": constant.StatusFailed, "error_msg": "1Panel restart causes failure"}).Error
return getTaskDb(t.WithByStatus(constant.StatusExecuting)).Model(&model.Task{}).Updates(map[string]interface{}{
"status": constant.StatusFailed, "error_msg": constant.InterruptedMsg, "end_at": time.Now(),
}).Error
}

func (t TaskRepo) CountExecutingTask() (int64, error) {
Expand Down
7 changes: 6 additions & 1 deletion agent/app/service/app.go
Original file line number Diff line number Diff line change
Expand Up @@ -550,14 +550,19 @@ func (a AppService) installWithHooks(req request.AppInstallCreate, executeScript

installTask, err := task.NewTaskWithOps(appInstall.Name, task.TaskInstall, task.TaskScopeApp, req.TaskID, appInstall.ID)
if err != nil {
appInstall.Status = constant.StatusInstallErr
appInstall.Message = err.Error()
if saveErr := appInstallRepo.Save(context.Background(), appInstall); saveErr != nil {
err = fmt.Errorf("%w; save failed install status: %v", err, saveErr)
}
return
}

if err = createLink(context.Background(), installTask, app, appInstall, req.Params); err != nil {
return
}

installApp := func(t *task.Task) error {
installApp := func(t *task.Task) (err error) {
if err = copyData(t, app, appDetail, appInstall, req); err != nil {
return err
}
Expand Down
41 changes: 37 additions & 4 deletions agent/app/service/app_install.go
Original file line number Diff line number Diff line change
Expand Up @@ -339,6 +339,11 @@ func operateAppWithLifecycleScripts(install model.AppInstall, req request.AppIns

operationTask, err := task.NewTaskWithOps(install.Name, taskType, task.TaskScopeApp, req.TaskID, install.ID)
if err != nil {
install.Status = constant.StatusUpErr
install.Message = err.Error()
if saveErr := appInstallRepo.Save(context.Background(), &install); saveErr != nil {
return fmt.Errorf("%w; save failed operation status: %v", err, saveErr)
}
return err
}
operation := string(req.Operate)
Expand Down Expand Up @@ -567,11 +572,15 @@ func (a *AppInstallService) SyncAll(systemInit bool) error {
return err
}
for _, i := range allList {
if i.Status == constant.StatusInstalling || i.Status == constant.StatusUpgrading || i.Status == constant.StatusRebuilding || i.Status == constant.StatusUninstalling {
if appInstallOperationInterruptedOnRestart(i.Status) {
if systemInit {
i.Status = constant.StatusError
i.Message = "1Panel restart causes the task to terminate"
_ = appInstallRepo.Save(context.Background(), &i)
i.Status = appInstallOperationFailureStatus(i.Status)
i.Message = constant.InterruptedMsg
if err := appInstallRepo.Save(context.Background(), &i); err != nil {
return err
}
} else if err := syncAppInstallStatus(&i, false); err != nil {
global.LOG.Errorf("sync install app[%s] error,mgs: %s", i.Name, err.Error())
}
continue
}
Expand Down Expand Up @@ -931,10 +940,34 @@ func (a *AppInstallService) GetParams(id uint) (*response.AppConfig, error) {
}

func syncAppInstallStatus(appInstall *model.AppInstall, force bool) error {
operation := ""
switch appInstall.Status {
case constant.StatusInstalling:
operation = task.TaskInstall
case constant.StatusUpgrading:
operation = task.TaskUpgrade
case constant.StatusRebuilding:
operation = task.TaskBuild
case constant.StatusUninstalling:
operation = task.TaskUninstall
case constant.StatusStarting, constant.StatusWaiting:
operation = task.TaskUpdate
case constant.StatusRestarting:
operation = task.TaskRestart
}
if operation != "" {
if err := reconcileAppInstallTaskFailure(appInstall, operation, global.DB, global.TaskDB); err != nil {
return err
}
}
switch appInstall.Status {
case constant.StatusInstalling, constant.StatusRebuilding, constant.StatusUpgrading, constant.StatusUninstalling,
constant.StatusStarting, constant.StatusRestarting, constant.StatusWaiting:
return nil
case constant.StatusInstallErr, constant.StatusUpgradeErr, constant.StatusUpErr:
if !force {
return nil
}
}
cli, err := docker.NewClient()
if err != nil {
Expand Down
76 changes: 76 additions & 0 deletions agent/app/service/app_operation_status.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,76 @@
package service

import (
"errors"

"github.com/1Panel-dev/1Panel/agent/app/model"
"github.com/1Panel-dev/1Panel/agent/constant"
"gorm.io/gorm"
)

func appInstallOperationFailureStatus(status string) string {
switch status {
case constant.StatusInstalling:
return constant.StatusInstallErr
case constant.StatusUpgrading:
return constant.StatusUpgradeErr
case constant.StatusUninstalling:
return constant.StatusError
default:
return constant.StatusUpErr
}
}

func appInstallOperationPending(status string) bool {
switch status {
case constant.StatusInstalling, constant.StatusRebuilding, constant.StatusUpgrading, constant.StatusUninstalling,
constant.StatusStarting, constant.StatusRestarting, constant.StatusWaiting:
return true
default:
return false
}
}

func appInstallOperationInterruptedOnRestart(status string) bool {
// ReStarting can also describe a container's Docker restart policy.
return appInstallOperationPending(status) && status != constant.StatusRestarting
}

func reconcileAppInstallTaskFailure(install *model.AppInstall, operation string, appDB, taskDB *gorm.DB) error {
if !appInstallOperationPending(install.Status) {
return nil
}
var operationTask model.Task
err := taskDB.Where("type = ? AND resource_id = ? AND operate = ?", "App", install.ID, operation).
Order("created_at DESC").First(&operationTask).Error
if errors.Is(err, gorm.ErrRecordNotFound) {
return nil
}
if err != nil {
return err
}
if operationTask.Status != constant.StatusFailed && operationTask.Status != constant.StatusCanceled {
return nil
}
// A previous operation must not turn a newly queued retry into a failure.
if operationTask.CreatedAt.Before(install.UpdatedAt) &&
(operationTask.EndAt.IsZero() || operationTask.EndAt.Before(install.UpdatedAt)) {
return nil
}
message := operationTask.ErrorMsg
if message == "" {
message = "the application operation task failed"
}
status := appInstallOperationFailureStatus(install.Status)
result := appDB.Model(&model.AppInstall{}).
Where("id = ? AND status = ? AND updated_at = ?", install.ID, install.Status, install.UpdatedAt).
Updates(map[string]interface{}{"status": status, "message": message})
if result.Error != nil {
return result.Error
}
if result.RowsAffected > 0 {
install.Status = status
install.Message = message
}
return nil
}
147 changes: 147 additions & 0 deletions agent/app/service/app_operation_status_test.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,147 @@
package service

import (
"path/filepath"
"testing"
"time"

"github.com/1Panel-dev/1Panel/agent/app/model"
"github.com/1Panel-dev/1Panel/agent/constant"
"github.com/glebarez/sqlite"
"gorm.io/gorm"
"gorm.io/gorm/clause"
"gorm.io/gorm/logger"
)

func operationStatusTestDB(t *testing.T) (*gorm.DB, *gorm.DB, model.AppInstall) {
t.Helper()
open := func(name string, schema interface{}) *gorm.DB {
db, err := gorm.Open(sqlite.Open(filepath.Join(t.TempDir(), name)), &gorm.Config{
Logger: logger.Default.LogMode(logger.Silent),
})
if err != nil {
t.Fatal(err)
}
if err := db.AutoMigrate(schema); err != nil {
t.Fatal(err)
}
t.Cleanup(func() {
sqlDB, _ := db.DB()
if sqlDB != nil {
_ = sqlDB.Close()
}
})
return db
}
appDB := open("apps.db", &model.AppInstall{})
taskDB := open("tasks.db", &model.Task{})
install := model.AppInstall{
Name: "gb10-ds-vision", AppId: 1, AppDetailId: 1,
Version: "nvidia-gb10-dspark-0.1.1", ContainerName: "vllm-dspark-1", ServiceName: "vllm-dspark",
Status: constant.StatusStarting, Env: `{"MODEL_DIR":"/models/vision"}`,
}
if err := appDB.Omit(clause.Associations).Create(&install).Error; err != nil {
t.Fatal(err)
}
return appDB, taskDB, install
}

func TestAppInstallFailedTaskRecoversPendingState(t *testing.T) {
for _, state := range []string{constant.StatusStarting, constant.StatusWaiting, constant.StatusRestarting, constant.StatusInstalling} {
t.Run(state, func(t *testing.T) {
appDB, taskDB, install := operationStatusTestDB(t)
install.Status = state
if err := appDB.Omit(clause.Associations).Save(&install).Error; err != nil {
t.Fatal(err)
}
task := model.Task{ID: "failed", Type: "App", Operate: "TaskUpdate", ResourceID: install.ID,
Status: constant.StatusFailed, ErrorMsg: "model startup script failed", CreatedAt: install.UpdatedAt.Add(time.Second), EndAt: install.UpdatedAt.Add(2 * time.Second)}
if err := taskDB.Create(&task).Error; err != nil {
t.Fatal(err)
}
if err := reconcileAppInstallTaskFailure(&install, task.Operate, appDB, taskDB); err != nil {
t.Fatal(err)
}
want := constant.StatusUpErr
if state == constant.StatusInstalling {
want = constant.StatusInstallErr
}
var stored model.AppInstall
if err := appDB.First(&stored, install.ID).Error; err != nil {
t.Fatal(err)
}
if install.Status != want || stored.Status != want || stored.Message != task.ErrorMsg || stored.Env != install.Env {
t.Fatalf("recovered state = %s/%s, message %q, env %q", install.Status, stored.Status, stored.Message, stored.Env)
}
})
}
}

func TestAppInstallPendingStateKeepsActiveAndUnrelatedTasks(t *testing.T) {
for _, taskStatus := range []string{constant.StatusExecuting, constant.StatusSuccess, "missing", "previous-failure", "other-resource"} {
t.Run(taskStatus, func(t *testing.T) {
appDB, taskDB, install := operationStatusTestDB(t)
task := model.Task{ID: "task", Type: "App", Operate: "TaskUpdate", ResourceID: install.ID,
Status: taskStatus, CreatedAt: install.UpdatedAt.Add(time.Second), EndAt: install.UpdatedAt.Add(2 * time.Second)}
if taskStatus == "previous-failure" {
task.Status = constant.StatusFailed
task.CreatedAt = install.UpdatedAt.Add(-2 * time.Hour)
task.EndAt = install.UpdatedAt.Add(-time.Hour)
}
if taskStatus == "other-resource" {
task.Status = constant.StatusFailed
task.ResourceID++
}
if taskStatus != "missing" {
if err := taskDB.Create(&task).Error; err != nil {
t.Fatal(err)
}
}
if err := reconcileAppInstallTaskFailure(&install, task.Operate, appDB, taskDB); err != nil {
t.Fatal(err)
}
if install.Status != constant.StatusStarting {
t.Fatalf("unrelated or active task changed pending state to %s", install.Status)
}
})
}
}

func TestAppInstallRecoveryDoesNotOverwriteConcurrentRetry(t *testing.T) {
appDB, taskDB, install := operationStatusTestDB(t)
task := model.Task{ID: "failed", Type: "App", Operate: "TaskUpdate", ResourceID: install.ID,
Status: constant.StatusCanceled, ErrorMsg: "interrupted", CreatedAt: install.UpdatedAt.Add(time.Second), EndAt: install.UpdatedAt.Add(2 * time.Second)}
if err := taskDB.Create(&task).Error; err != nil {
t.Fatal(err)
}
if err := appDB.Callback().Update().Before("gorm:update").Register("new-operation", func(db *gorm.DB) {
if err := db.Exec("UPDATE app_installs SET updated_at = ?, message = ? WHERE id = ?", install.UpdatedAt.Add(time.Hour), "new attempt", install.ID).Error; err != nil {
t.Error(err)
}
}); err != nil {
t.Fatal(err)
}
if err := reconcileAppInstallTaskFailure(&install, task.Operate, appDB, taskDB); err != nil {
t.Fatal(err)
}
var stored model.AppInstall
if err := appDB.First(&stored, install.ID).Error; err != nil {
t.Fatal(err)
}
if stored.Status != constant.StatusStarting || stored.Message != "new attempt" {
t.Fatalf("recovery overwrote a newer operation: %s, %s", stored.Status, stored.Message)
}
}

func TestAppInstallRestartRecoveryCoversLifecycleStates(t *testing.T) {
for _, state := range []string{constant.StatusStarting, constant.StatusWaiting, constant.StatusInstalling} {
if !appInstallOperationInterruptedOnRestart(state) {
t.Errorf("restart recovery omitted %s", state)
}
}
for _, state := range []string{constant.StatusRunning, constant.StatusStopped, constant.StatusRestarting} {
if appInstallOperationInterruptedOnRestart(state) {
t.Errorf("restart recovery changes a Docker runtime state: %s", state)
}
}
}
10 changes: 7 additions & 3 deletions agent/app/service/app_utils.go
Original file line number Diff line number Diff line change
Expand Up @@ -1406,9 +1406,8 @@ func handleErr(install model.AppInstall, err error, out string) error {
}

func doNotNeedSync(installed model.AppInstall) bool {
return installed.Status == constant.StatusInstalling || installed.Status == constant.StatusRebuilding || installed.Status == constant.StatusUpgrading ||
installed.Status == constant.StatusSyncing || installed.Status == constant.StatusUninstalling || installed.Status == constant.StatusInstallErr ||
installed.Status == constant.StatusStarting || installed.Status == constant.StatusRestarting || installed.Status == constant.StatusWaiting
return appInstallOperationPending(installed.Status) || installed.Status == constant.StatusSyncing ||
installed.Status == constant.StatusInstallErr || installed.Status == constant.StatusUpgradeErr || installed.Status == constant.StatusUpErr
}

func synAppInstall(containers map[string]container.Summary, appInstall *model.AppInstall, force bool) {
Expand Down Expand Up @@ -1502,6 +1501,11 @@ func handleInstalled(appInstallList []model.AppInstall, updated, sync, checkUpda
}

for _, installed := range appInstallList {
if sync && appInstallOperationPending(installed.Status) {
if err := syncAppInstallStatus(&installed, false); err != nil {
return nil, err
}
}
if updated && ignoreUpdate(installed) {
continue
}
Expand Down
Loading
Loading