feat(worker): add durable background jobs

This commit is contained in:
2026-06-05 15:38:03 +08:00
parent 52b6497d3a
commit 2cb02c8a91
8 changed files with 732 additions and 24 deletions
+1
View File
@@ -242,6 +242,7 @@ func autoMigrate(db *gorm.DB) error {
&model.EmailTemplate{},
&model.MessageReaction{},
&model.Report{},
&model.BackgroundJob{},
// S6: WorkingHour — out-of-office / business hours per inbox
&model.WorkingHour{},
}
+41
View File
@@ -0,0 +1,41 @@
package model
import (
"encoding/json"
"time"
)
const (
BackgroundJobStatusQueued = "queued"
BackgroundJobStatusRunning = "running"
BackgroundJobStatusRetrying = "retrying"
BackgroundJobStatusCompleted = "completed"
BackgroundJobStatusDead = "dead"
)
const DefaultBackgroundJobQueue = "default"
// BackgroundJob stores durable background work in the database.
// Reference: Chatwoot ActiveJob/Sidekiq jobs with retry, delayed scheduling,
// mutex/idempotency, and observable failure state.
type BackgroundJob struct {
ID uint `gorm:"primaryKey" json:"id"`
Queue string `gorm:"size:100;not null;default:'default';index:idx_background_jobs_ready,priority:2" json:"queue"`
JobType string `gorm:"size:150;not null;index" json:"job_type"`
Payload json.RawMessage `gorm:"type:jsonb;not null;default:'{}'" json:"payload"`
Status string `gorm:"size:50;not null;default:'queued';index:idx_background_jobs_ready,priority:1" json:"status"`
Priority int `gorm:"not null;default:0" json:"priority"`
Attempts int `gorm:"not null;default:0" json:"attempts"`
MaxAttempts int `gorm:"not null;default:3" json:"max_attempts"`
ScheduledAt time.Time `gorm:"not null;index:idx_background_jobs_ready,priority:3" json:"scheduled_at"`
LockedAt *time.Time `gorm:"index" json:"locked_at,omitempty"`
LockedBy string `gorm:"size:150" json:"locked_by,omitempty"`
IdempotencyKey string `gorm:"size:255;index" json:"idempotency_key,omitempty"`
LastError string `gorm:"type:text" json:"last_error,omitempty"`
FinishedAt *time.Time `json:"finished_at,omitempty"`
FailedAt *time.Time `json:"failed_at,omitempty"`
CreatedAt time.Time `gorm:"autoCreateTime" json:"created_at"`
UpdatedAt time.Time `gorm:"autoUpdateTime" json:"updated_at"`
}
func (BackgroundJob) TableName() string { return "background_jobs" }
+1
View File
@@ -168,6 +168,7 @@ func defaultTestModels() []interface{} {
&model.Mention{},
&model.IntegrationHook{},
&model.IntegrationApp{},
&model.BackgroundJob{},
}
}
+407 -17
View File
@@ -1,31 +1,421 @@
package worker
// Worker processes background jobs such as:
// - Auto-assignment of conversations to agents
// - Sending outgoing messages through channels
// - Processing webhook events asynchronously
// - Notification delivery
import (
"context"
"encoding/json"
"errors"
"fmt"
"math"
"sync"
"time"
// WorkerPool manages a set of background job processors.
// This is a placeholder — will be implemented with proper job queue
// (e.g., Redis-based or database-backed queue).
"github.com/gochat/gochat/internal/model"
"gorm.io/gorm"
"gorm.io/gorm/clause"
)
var ErrWorkerDatabaseRequired = errors.New("worker database is required")
// JobHandler performs one durable background job.
type JobHandler func(context.Context, *model.BackgroundJob) error
type BackoffFunc func(attempt int) time.Duration
// WorkerPool manages durable background job processors.
// Reference: Chatwoot ActiveJob/Sidekiq workers use persistent jobs, scheduled
// execution, retry, and dead-letter visibility for async side effects.
type WorkerPool struct {
// TODO: implement with job queue
db *gorm.DB
handlers map[string]JobHandler
queues []string
workerID string
workerCount int
pollInterval time.Duration
staleLockTimeout time.Duration
backoff BackoffFunc
now func() time.Time
mu sync.RWMutex
ctx context.Context
cancel context.CancelFunc
wg sync.WaitGroup
}
// NewWorkerPool creates a new WorkerPool instance.
func NewWorkerPool() *WorkerPool {
return &WorkerPool{}
type Option func(*WorkerPool)
// NewWorkerPool creates a WorkerPool. The variadic DB preserves older no-arg
// construction while allowing Phase 5 durable workers to be wired incrementally.
func NewWorkerPool(db ...*gorm.DB) *WorkerPool {
wp := &WorkerPool{
handlers: make(map[string]JobHandler),
queues: []string{model.DefaultBackgroundJobQueue},
workerID: fmt.Sprintf("worker-%d", time.Now().UnixNano()),
workerCount: 1,
pollInterval: 500 * time.Millisecond,
staleLockTimeout: 15 * time.Minute,
backoff: defaultBackoff,
now: time.Now,
}
if len(db) > 0 {
wp.db = db[0]
}
return wp
}
func NewWorkerPoolWithOptions(db *gorm.DB, opts ...Option) *WorkerPool {
wp := NewWorkerPool(db)
for _, opt := range opts {
opt(wp)
}
return wp
}
func WithWorkerCount(count int) Option {
return func(wp *WorkerPool) {
if count > 0 {
wp.workerCount = count
}
}
}
func WithQueues(queues ...string) Option {
return func(wp *WorkerPool) {
if len(queues) > 0 {
wp.queues = append([]string(nil), queues...)
}
}
}
func WithPollInterval(interval time.Duration) Option {
return func(wp *WorkerPool) {
if interval > 0 {
wp.pollInterval = interval
}
}
}
func WithWorkerID(id string) Option {
return func(wp *WorkerPool) {
if id != "" {
wp.workerID = id
}
}
}
func WithStaleLockTimeout(timeout time.Duration) Option {
return func(wp *WorkerPool) {
if timeout > 0 {
wp.staleLockTimeout = timeout
}
}
}
func WithBackoff(backoff BackoffFunc) Option {
return func(wp *WorkerPool) {
if backoff != nil {
wp.backoff = backoff
}
}
}
func WithNow(now func() time.Time) Option {
return func(wp *WorkerPool) {
if now != nil {
wp.now = now
}
}
}
func (wp *WorkerPool) Register(jobType string, handler JobHandler) {
wp.mu.Lock()
defer wp.mu.Unlock()
wp.handlers[jobType] = handler
}
type EnqueueOption func(*model.BackgroundJob)
func WithQueue(queue string) EnqueueOption {
return func(job *model.BackgroundJob) {
if queue != "" {
job.Queue = queue
}
}
}
func WithScheduledAt(at time.Time) EnqueueOption {
return func(job *model.BackgroundJob) {
if !at.IsZero() {
job.ScheduledAt = at
}
}
}
func WithMaxAttempts(max int) EnqueueOption {
return func(job *model.BackgroundJob) {
if max > 0 {
job.MaxAttempts = max
}
}
}
func WithPriority(priority int) EnqueueOption {
return func(job *model.BackgroundJob) {
job.Priority = priority
}
}
func WithIdempotencyKey(key string) EnqueueOption {
return func(job *model.BackgroundJob) {
job.IdempotencyKey = key
}
}
func (wp *WorkerPool) Enqueue(ctx context.Context, jobType string, payload any, opts ...EnqueueOption) (*model.BackgroundJob, error) {
if wp.db == nil {
return nil, ErrWorkerDatabaseRequired
}
if jobType == "" {
return nil, errors.New("job type is required")
}
payloadBytes, err := marshalPayload(payload)
if err != nil {
return nil, err
}
job := &model.BackgroundJob{
Queue: model.DefaultBackgroundJobQueue,
JobType: jobType,
Payload: payloadBytes,
Status: model.BackgroundJobStatusQueued,
MaxAttempts: 3,
ScheduledAt: wp.now(),
}
for _, opt := range opts {
opt(job)
}
if job.ScheduledAt.IsZero() {
job.ScheduledAt = wp.now()
}
if job.MaxAttempts <= 0 {
job.MaxAttempts = 3
}
if job.IdempotencyKey != "" {
var existing model.BackgroundJob
err := wp.db.WithContext(ctx).Where("idempotency_key = ?", job.IdempotencyKey).First(&existing).Error
if err == nil {
return &existing, nil
}
if !errors.Is(err, gorm.ErrRecordNotFound) {
return nil, err
}
}
if err := wp.db.WithContext(ctx).Create(job).Error; err != nil {
if job.IdempotencyKey != "" {
var existing model.BackgroundJob
if findErr := wp.db.WithContext(ctx).Where("idempotency_key = ?", job.IdempotencyKey).First(&existing).Error; findErr == nil {
return &existing, nil
}
}
return nil, err
}
return job, nil
}
// Start initializes and begins processing background jobs.
func (wp *WorkerPool) Start() error {
// TODO: implement
if wp.db == nil {
return nil
}
wp.mu.Lock()
if wp.cancel != nil {
wp.mu.Unlock()
return nil
}
wp.ctx, wp.cancel = context.WithCancel(context.Background())
ctx := wp.ctx
workerCount := wp.workerCount
wp.mu.Unlock()
if _, err := wp.RequeueStaleJobs(ctx); err != nil {
return err
}
for i := 0; i < workerCount; i++ {
wp.wg.Add(1)
go wp.run(ctx)
}
return nil
}
// Stop gracefully shuts down the worker pool.
func (wp *WorkerPool) Stop() error {
// TODO: implement
wp.mu.Lock()
cancel := wp.cancel
wp.cancel = nil
wp.ctx = nil
wp.mu.Unlock()
if cancel != nil {
cancel()
}
wp.wg.Wait()
return nil
}
}
func (wp *WorkerPool) ProcessOne(ctx context.Context) (bool, error) {
if wp.db == nil {
return false, ErrWorkerDatabaseRequired
}
job, err := wp.claimNext(ctx)
if err != nil {
if errors.Is(err, gorm.ErrRecordNotFound) {
return false, nil
}
return false, err
}
return true, wp.perform(ctx, job)
}
func (wp *WorkerPool) RequeueStaleJobs(ctx context.Context) (int64, error) {
if wp.db == nil {
return 0, ErrWorkerDatabaseRequired
}
cutoff := wp.now().Add(-wp.staleLockTimeout)
updates := map[string]any{
"status": model.BackgroundJobStatusRetrying,
"locked_at": nil,
"locked_by": "",
}
result := wp.db.WithContext(ctx).Model(&model.BackgroundJob{}).
Where("status = ? AND locked_at IS NOT NULL AND locked_at < ?", model.BackgroundJobStatusRunning, cutoff).
Updates(updates)
return result.RowsAffected, result.Error
}
func (wp *WorkerPool) run(ctx context.Context) {
defer wp.wg.Done()
ticker := time.NewTicker(wp.pollInterval)
defer ticker.Stop()
for {
processed, _ := wp.ProcessOne(ctx)
if processed {
continue
}
select {
case <-ctx.Done():
return
case <-ticker.C:
}
}
}
func (wp *WorkerPool) claimNext(ctx context.Context) (*model.BackgroundJob, error) {
var job model.BackgroundJob
err := wp.db.WithContext(ctx).Transaction(func(tx *gorm.DB) error {
query := tx.Where("status IN ? AND scheduled_at <= ?", []string{model.BackgroundJobStatusQueued, model.BackgroundJobStatusRetrying}, wp.now())
if len(wp.queues) > 0 {
query = query.Where("queue IN ?", wp.queues)
}
if tx.Dialector != nil && tx.Dialector.Name() == "postgres" {
query = query.Clauses(clause.Locking{Strength: "UPDATE", Options: "SKIP LOCKED"})
}
if err := query.Order("priority DESC, scheduled_at ASC, id ASC").First(&job).Error; err != nil {
return err
}
lockedAt := wp.now()
updates := map[string]any{
"status": model.BackgroundJobStatusRunning,
"locked_at": &lockedAt,
"locked_by": wp.workerID,
"attempts": gorm.Expr("attempts + 1"),
}
if err := tx.Model(&model.BackgroundJob{}).Where("id = ?", job.ID).Updates(updates).Error; err != nil {
return err
}
return tx.First(&job, job.ID).Error
})
if err != nil {
return nil, err
}
return &job, nil
}
func (wp *WorkerPool) perform(ctx context.Context, job *model.BackgroundJob) error {
handler := wp.handlerFor(job.JobType)
if handler == nil {
return wp.fail(ctx, job, fmt.Errorf("no handler registered for job type %q", job.JobType))
}
if err := handler(ctx, job); err != nil {
return wp.fail(ctx, job, err)
}
finishedAt := wp.now()
updates := map[string]any{
"status": model.BackgroundJobStatusCompleted,
"locked_at": nil,
"locked_by": "",
"last_error": "",
"finished_at": &finishedAt,
}
return wp.db.WithContext(ctx).Model(&model.BackgroundJob{}).Where("id = ?", job.ID).Updates(updates).Error
}
func (wp *WorkerPool) fail(ctx context.Context, job *model.BackgroundJob, err error) error {
now := wp.now()
updates := map[string]any{
"locked_at": nil,
"locked_by": "",
"last_error": err.Error(),
}
if job.Attempts >= job.MaxAttempts {
updates["status"] = model.BackgroundJobStatusDead
updates["failed_at"] = &now
} else {
updates["status"] = model.BackgroundJobStatusRetrying
updates["scheduled_at"] = now.Add(wp.backoff(job.Attempts))
}
if updateErr := wp.db.WithContext(ctx).Model(&model.BackgroundJob{}).Where("id = ?", job.ID).Updates(updates).Error; updateErr != nil {
return updateErr
}
return err
}
func (wp *WorkerPool) handlerFor(jobType string) JobHandler {
wp.mu.RLock()
defer wp.mu.RUnlock()
return wp.handlers[jobType]
}
func marshalPayload(payload any) (json.RawMessage, error) {
if payload == nil {
return json.RawMessage(`{}`), nil
}
switch value := payload.(type) {
case json.RawMessage:
if !json.Valid(value) {
return nil, errors.New("job payload must be valid JSON")
}
return value, nil
case []byte:
if !json.Valid(value) {
return nil, errors.New("job payload must be valid JSON")
}
return json.RawMessage(value), nil
default:
payloadBytes, err := json.Marshal(value)
if err != nil {
return nil, err
}
return payloadBytes, nil
}
}
func defaultBackoff(attempt int) time.Duration {
if attempt <= 0 {
attempt = 1
}
seconds := math.Pow(2, float64(attempt-1))
if seconds > 300 {
seconds = 300
}
return time.Duration(seconds) * time.Second
}
+221
View File
@@ -0,0 +1,221 @@
package worker
import (
"context"
"encoding/json"
"errors"
"sync/atomic"
"testing"
"time"
"github.com/gochat/gochat/internal/model"
"gorm.io/driver/sqlite"
"gorm.io/gorm"
"gorm.io/gorm/logger"
)
func newWorkerTestDB(t *testing.T) *gorm.DB {
t.Helper()
db, err := gorm.Open(sqlite.Open("file:worker-test?mode=memory&cache=shared"), &gorm.Config{Logger: logger.Default.LogMode(logger.Silent)})
if err != nil {
t.Fatalf("open sqlite: %v", err)
}
sqlDB, err := db.DB()
if err != nil {
t.Fatalf("sqlite db handle: %v", err)
}
sqlDB.SetMaxOpenConns(1)
if err := db.AutoMigrate(&model.BackgroundJob{}); err != nil {
t.Fatalf("migrate background jobs: %v", err)
}
t.Cleanup(func() {
db.Exec("DELETE FROM background_jobs")
sqlDB.Close()
})
return db
}
func loadJob(t *testing.T, db *gorm.DB, id uint) model.BackgroundJob {
t.Helper()
var job model.BackgroundJob
if err := db.First(&job, id).Error; err != nil {
t.Fatalf("load job: %v", err)
}
return job
}
func TestWorkerPoolEnqueueStoresPayloadAndIdempotency(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }))
job, err := wp.Enqueue(context.Background(), "contact_export", map[string]any{"account_id": 1}, WithIdempotencyKey("contact-export:1"), WithQueue("exports"), WithMaxAttempts(5), WithPriority(10))
if err != nil {
t.Fatalf("enqueue: %v", err)
}
duplicate, err := wp.Enqueue(context.Background(), "contact_export", map[string]any{"account_id": 2}, WithIdempotencyKey("contact-export:1"))
if err != nil {
t.Fatalf("enqueue duplicate: %v", err)
}
if duplicate.ID != job.ID {
t.Fatalf("expected duplicate enqueue to return existing job %d, got %d", job.ID, duplicate.ID)
}
reloaded := loadJob(t, db, job.ID)
if reloaded.Queue != "exports" || reloaded.JobType != "contact_export" || reloaded.Status != model.BackgroundJobStatusQueued || reloaded.MaxAttempts != 5 || reloaded.Priority != 10 {
t.Fatalf("unexpected job fields: %+v", reloaded)
}
var payload map[string]any
if err := json.Unmarshal(reloaded.Payload, &payload); err != nil {
t.Fatalf("unmarshal payload: %v", err)
}
if payload["account_id"].(float64) != 1 {
t.Fatalf("unexpected payload: %s", string(reloaded.Payload))
}
}
func TestWorkerPoolProcessOneCompletesDueJob(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }), WithWorkerID("test-worker"))
var handled atomic.Int32
wp.Register("send_reply", func(ctx context.Context, job *model.BackgroundJob) error {
handled.Add(1)
if job.Attempts != 1 || job.LockedBy != "test-worker" || job.LockedAt == nil {
t.Fatalf("job was not claimed before handler: %+v", job)
}
return nil
})
job, err := wp.Enqueue(context.Background(), "send_reply", map[string]any{"message_id": 7})
if err != nil {
t.Fatalf("enqueue: %v", err)
}
processed, err := wp.ProcessOne(context.Background())
if err != nil {
t.Fatalf("process one: %v", err)
}
if !processed || handled.Load() != 1 {
t.Fatalf("expected one handled job, processed=%v handled=%d", processed, handled.Load())
}
reloaded := loadJob(t, db, job.ID)
if reloaded.Status != model.BackgroundJobStatusCompleted || reloaded.FinishedAt == nil || reloaded.LockedAt != nil || reloaded.LockedBy != "" {
t.Fatalf("expected completed unlocked job: %+v", reloaded)
}
}
func TestWorkerPoolRetriesThenDeadLettersFailures(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }), WithBackoff(func(attempt int) time.Duration { return 0 }))
boom := errors.New("provider timeout")
wp.Register("webhook_delivery", func(ctx context.Context, job *model.BackgroundJob) error { return boom })
job, err := wp.Enqueue(context.Background(), "webhook_delivery", nil, WithMaxAttempts(2))
if err != nil {
t.Fatalf("enqueue: %v", err)
}
processed, err := wp.ProcessOne(context.Background())
if !processed || !errors.Is(err, boom) {
t.Fatalf("expected first failure, processed=%v err=%v", processed, err)
}
reloaded := loadJob(t, db, job.ID)
if reloaded.Status != model.BackgroundJobStatusRetrying || reloaded.Attempts != 1 || reloaded.LastError != boom.Error() || reloaded.FailedAt != nil {
t.Fatalf("expected retrying job after first failure: %+v", reloaded)
}
processed, err = wp.ProcessOne(context.Background())
if !processed || !errors.Is(err, boom) {
t.Fatalf("expected second failure, processed=%v err=%v", processed, err)
}
reloaded = loadJob(t, db, job.ID)
if reloaded.Status != model.BackgroundJobStatusDead || reloaded.Attempts != 2 || reloaded.FailedAt == nil || reloaded.LockedAt != nil {
t.Fatalf("expected dead-lettered job: %+v", reloaded)
}
}
func TestWorkerPoolRespectsScheduleAndQueues(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }), WithQueues("critical"))
wp.Register("sla_scan", func(ctx context.Context, job *model.BackgroundJob) error { return nil })
if _, err := wp.Enqueue(context.Background(), "sla_scan", nil, WithQueue("default")); err != nil {
t.Fatalf("enqueue default: %v", err)
}
if _, err := wp.Enqueue(context.Background(), "sla_scan", nil, WithQueue("critical"), WithScheduledAt(now.Add(time.Hour))); err != nil {
t.Fatalf("enqueue future: %v", err)
}
processed, err := wp.ProcessOne(context.Background())
if err != nil || processed {
t.Fatalf("expected no eligible job, processed=%v err=%v", processed, err)
}
}
func TestWorkerPoolRequeuesStaleRunningJobs(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }), WithStaleLockTimeout(time.Minute))
lockedAt := now.Add(-2 * time.Minute)
job := model.BackgroundJob{
Queue: model.DefaultBackgroundJobQueue,
JobType: "captain_document_sync",
Payload: json.RawMessage(`{}`),
Status: model.BackgroundJobStatusRunning,
MaxAttempts: 3,
ScheduledAt: now.Add(-time.Hour),
LockedAt: &lockedAt,
LockedBy: "dead-worker",
}
if err := db.Create(&job).Error; err != nil {
t.Fatalf("create stale job: %v", err)
}
count, err := wp.RequeueStaleJobs(context.Background())
if err != nil {
t.Fatalf("requeue stale: %v", err)
}
if count != 1 {
t.Fatalf("expected 1 stale job requeued, got %d", count)
}
reloaded := loadJob(t, db, job.ID)
if reloaded.Status != model.BackgroundJobStatusRetrying || reloaded.LockedAt != nil || reloaded.LockedBy != "" {
t.Fatalf("expected retrying unlocked stale job: %+v", reloaded)
}
}
func TestWorkerPoolStartAndStopProcessJobs(t *testing.T) {
db := newWorkerTestDB(t)
now := time.Date(2026, 6, 5, 10, 0, 0, 0, time.UTC)
wp := NewWorkerPoolWithOptions(db, WithNow(func() time.Time { return now }), WithPollInterval(5*time.Millisecond))
var handled atomic.Int32
wp.Register("event_dispatch", func(ctx context.Context, job *model.BackgroundJob) error {
handled.Add(1)
return nil
})
job, err := wp.Enqueue(context.Background(), "event_dispatch", map[string]any{"event": "conversation_created"})
if err != nil {
t.Fatalf("enqueue: %v", err)
}
if err := wp.Start(); err != nil {
t.Fatalf("start worker: %v", err)
}
defer wp.Stop()
deadline := time.Now().Add(time.Second)
for time.Now().Before(deadline) {
if handled.Load() == 1 {
break
}
time.Sleep(10 * time.Millisecond)
}
if handled.Load() != 1 {
t.Fatalf("worker loop did not process job")
}
if err := wp.Stop(); err != nil {
t.Fatalf("stop worker: %v", err)
}
reloaded := loadJob(t, db, job.ID)
if reloaded.Status != model.BackgroundJobStatusCompleted {
t.Fatalf("expected completed job after worker loop: %+v", reloaded)
}
}