feat: stop zombie and endless tasks

models/bots/task.go

@@ -38,7 +38,7 @@ type Task struct {
 	RunnerID int64 `xorm:"index"`
 	Result   runnerv1.Result
 	Status   Status             `xorm:"index"`
-	Started  timeutil.TimeStamp
+	Started  timeutil.TimeStamp `xorm:"index"`
 	Stopped  timeutil.TimeStamp
 
 	Token          string `xorm:"-"`
@@ -54,7 +54,7 @@ type Task struct {
 	LogExpired   bool        // files that are too old will be deleted
 
 	Created timeutil.TimeStamp `xorm:"created"`
-	Updated timeutil.TimeStamp `xorm:"updated"`
+	Updated timeutil.TimeStamp `xorm:"updated index"`
 }
 
 func init() {
@@ -412,6 +412,57 @@ func UpdateTaskByState(state *runnerv1.TaskState) (*Task, error) {
 	return task, nil
 }
 
+func StopTask(ctx context.Context, task *Task, result runnerv1.Result) (*Task, error) {
+	ctx, commiter, err := db.TxContext()
+	if err != nil {
+		return nil, err
+	}
+	defer commiter.Close()
+
+	e := db.GetEngine(ctx)
+
+	now := timeutil.TimeStampNow()
+	task.Result = result
+	if task.Result != runnerv1.Result_RESULT_UNSPECIFIED {
+		task.Status = Status(task.Result)
+		task.Stopped = now
+		if _, err := UpdateRunJob(ctx, &RunJob{
+			ID:      task.JobID,
+			Status:  task.Status,
+			Stopped: task.Stopped,
+		}, nil); err != nil {
+			return nil, err
+		}
+	}
+
+	if _, err := e.ID(task.ID).Update(task); err != nil {
+		return nil, err
+	}
+
+	if err := task.LoadAttributes(ctx); err != nil {
+		return nil, err
+	}
+
+	for _, step := range task.Steps {
+		if step.Result == runnerv1.Result_RESULT_UNSPECIFIED {
+			step.Result = result
+			if step.Started == 0 {
+				step.Started = now
+			}
+			step.Stopped = now
+		}
+		if _, err := e.ID(step.ID).Update(step); err != nil {
+			return nil, err
+		}
+	}
+
+	if err := commiter.Commit(); err != nil {
+		return nil, err
+	}
+
+	return task, nil
+}
+
 func isSubset(set, subset []string) bool {
 	m := make(map[string]struct{}, len(set))
 	for _, v := range set {

models/bots/task_list.go

@@ -0,0 +1,50 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package bots
+
+import (
+	"context"
+
+	"code.gitea.io/gitea/models/db"
+	"code.gitea.io/gitea/modules/timeutil"
+	"xorm.io/builder"
+)
+
+type TaskList []*Task
+
+type FindTaskOptions struct {
+	db.ListOptions
+	Status        Status
+	UpdatedBefore timeutil.TimeStamp
+	StartedBefore timeutil.TimeStamp
+}
+
+func (opts FindTaskOptions) toConds() builder.Cond {
+	cond := builder.NewCond()
+	if opts.Status > StatusUnknown {
+		cond = cond.And(builder.Eq{"status": opts.Status})
+	}
+	if opts.UpdatedBefore > 0 {
+		cond = cond.And(builder.Lt{"updated": opts.UpdatedBefore})
+	}
+	if opts.StartedBefore > 0 {
+		cond = cond.And(builder.Lt{"started": opts.StartedBefore})
+	}
+	return cond
+}
+
+func FindTasks(ctx context.Context, opts FindTaskOptions) (TaskList, int64, error) {
+	e := db.GetEngine(ctx).Where(opts.toConds())
+	if opts.PageSize > 0 && opts.Page >= 1 {
+		e.Limit(opts.PageSize, (opts.Page-1)*opts.PageSize)
+	}
+	var tasks TaskList
+	total, err := e.FindAndCount(&tasks)
+	return tasks, total, err
+}
+
+func CountTasks(ctx context.Context, opts FindTaskOptions) (int64, error) {
+	return db.GetEngine(ctx).Where(opts.toConds()).Count(new(Task))
+}

options/locale/locale_en-US.ini

@@ -2574,6 +2574,9 @@ dashboard.delete_old_actions = Delete all old actions from database
 dashboard.delete_old_actions.started = Delete all old actions from database started.
 dashboard.update_checker = Update checker
 dashboard.delete_old_system_notices = Delete all old system notices from database
+dashboard.stop_zombie_tasks = Stop zombie tasks
+dashboard.stop_endless_tasks = Stop endless tasks
+dashboard.cancel_abandoned_jobs = Cancel abandoned jobs
 
 users.user_manage_panel = User Account Management
 users.new_account = Create User Account

services/bots/clear_tasks.go

@@ -0,0 +1,67 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package bots
+
+import (
+	"context"
+	"time"
+
+	bots_model "code.gitea.io/gitea/models/bots"
+	"code.gitea.io/gitea/modules/log"
+	"code.gitea.io/gitea/modules/timeutil"
+	runnerv1 "gitea.com/gitea/proto-go/runner/v1"
+)
+
+const (
+	zombieTaskTimeout   = 10 * time.Minute
+	endlessTaskTimeout  = 3 * time.Hour  // the task is running for a long time with updates
+	abandonedJobTimeout = 24 * time.Hour // the job is waiting for being picked by a runner
+)
+
+// StopZombieTasks stops the task which have running status, but haven't been updated for a long time
+func StopZombieTasks(ctx context.Context) error {
+	tasks, _, err := bots_model.FindTasks(ctx, bots_model.FindTaskOptions{
+		Status:        bots_model.StatusRunning,
+		UpdatedBefore: timeutil.TimeStamp(time.Now().Add(-zombieTaskTimeout).Unix()),
+	})
+	if err != nil {
+		log.Warn("find zombie tasks: %v", err)
+		return err
+	}
+
+	for _, task := range tasks {
+		if _, err := bots_model.StopTask(ctx, task, runnerv1.Result_RESULT_FAILURE); err != nil {
+			log.Warn("stop zombie task %v: %v", task.ID, err)
+			// go on
+		}
+	}
+	return nil
+}
+
+// StopEndlessTasks stops the tasks which have running status and continuous updates, but don't end for a long time
+func StopEndlessTasks(ctx context.Context) error {
+	tasks, _, err := bots_model.FindTasks(ctx, bots_model.FindTaskOptions{
+		Status:        bots_model.StatusRunning,
+		StartedBefore: timeutil.TimeStamp(time.Now().Add(-endlessTaskTimeout).Unix()),
+	})
+	if err != nil {
+		log.Warn("find endless tasks: %v", err)
+		return err
+	}
+
+	for _, task := range tasks {
+		if _, err := bots_model.StopTask(ctx, task, runnerv1.Result_RESULT_FAILURE); err != nil {
+			log.Warn("stop endless task %v: %v", task.ID, err)
+			// go on
+		}
+	}
+	return nil
+}
+
+// CancelAbandonedJobs cancels the jobs which have waiting status, but haven't been picked by a runner for a long time
+func CancelAbandonedJobs(ctx context.Context) error {
+	// TODO
+	return nil
+}

services/cron/cron.go

@@ -31,6 +31,7 @@ func NewContext(original context.Context) {
 	_, _, finished := process.GetManager().AddTypedContext(graceful.GetManager().ShutdownContext(), "Service: Cron", process.SystemProcessType, true)
 	initBasicTasks()
 	initExtendedTasks()
+	initBotsTasks()
 
 	lock.Lock()
 	for _, task := range tasks {

services/cron/tasks_bots.go

@@ -0,0 +1,48 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package cron
+
+import (
+	"context"
+
+	user_model "code.gitea.io/gitea/models/user"
+	bots_service "code.gitea.io/gitea/services/bots"
+)
+
+func initBotsTasks() {
+	registerStopZombieTasks()
+	registerStopEndlessTasks()
+	registerCancelAbandonedJobs()
+}
+
+func registerStopZombieTasks() {
+	RegisterTaskFatal("stop_zombie_tasks", &BaseConfig{
+		Enabled:    true,
+		RunAtStart: true,
+		Schedule:   "@every 5m",
+	}, func(ctx context.Context, _ *user_model.User, cfg Config) error {
+		return bots_service.StopZombieTasks(ctx)
+	})
+}
+
+func registerStopEndlessTasks() {
+	RegisterTaskFatal("stop_endless_tasks", &BaseConfig{
+		Enabled:    true,
+		RunAtStart: true,
+		Schedule:   "@every 30m",
+	}, func(ctx context.Context, _ *user_model.User, cfg Config) error {
+		return bots_service.StopEndlessTasks(ctx)
+	})
+}
+
+func registerCancelAbandonedJobs() {
+	RegisterTaskFatal("cancel_abandoned_jobs", &BaseConfig{
+		Enabled:    true,
+		RunAtStart: true,
+		Schedule:   "@every 6h",
+	}, func(ctx context.Context, _ *user_model.User, cfg Config) error {
+		return bots_service.CancelAbandonedJobs(ctx)
+	})
+}