feat(jobs): worker pool with retry/dead-letter/panic-recovery

2026-05-05 19:57:23 +08:00
parent f235b1be63
commit 1a83a5b1cf
2 changed files with 316 additions and 0 deletions
@@ -0,0 +1,168 @@
+//go:build integration
+
+package jobs_test
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/yan1h/agent-coding-workflow/internal/infra/clock"
+	"github.com/yan1h/agent-coding-workflow/internal/jobs"
+)
+
+type recordingHandler struct {
+	typ       jobs.JobType
+	calls     atomic.Int32
+	wantErr   error
+	beforeRet func()
+}
+
+func (h *recordingHandler) Type() jobs.JobType { return h.typ }
+func (h *recordingHandler) Handle(ctx context.Context, job jobs.Job) error {
+	h.calls.Add(1)
+	if h.beforeRet != nil {
+		h.beforeRet()
+	}
+	return h.wantErr
+}
+
+func TestWorker_HandlesAndCompletes(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	_, err := repo.Enqueue(ctx, jobs.TypeWebhookDeliver, json.RawMessage(`{}`), time.Now(), 5)
+	require.NoError(t, err)
+
+	h := &recordingHandler{typ: jobs.TypeWebhookDeliver}
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{jobs.TypeWebhookDeliver: h},
+		jobs.WorkerOptions{
+			ID:           "test#1",
+			PollInterval: 5 * time.Millisecond,
+			Backoff:      []time.Duration{30 * time.Second},
+			Clock:        clock.Real(),
+			Log:          discardLogger(),
+		})
+	cctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool { return h.calls.Load() >= 1 }, 200*time.Millisecond, 5*time.Millisecond)
+	cancel()
+	w.Wait()
+}
+
+func TestWorker_RetryableErrorReschedules(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	enqueued, err := repo.Enqueue(ctx, jobs.TypeWebhookDeliver, json.RawMessage(`{}`), time.Now(), 5)
+	require.NoError(t, err)
+
+	h := &recordingHandler{typ: jobs.TypeWebhookDeliver, wantErr: errors.New("transient")}
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{jobs.TypeWebhookDeliver: h},
+		jobs.WorkerOptions{
+			ID:           "t#1",
+			PollInterval: 5 * time.Millisecond,
+			Backoff:      []time.Duration{30 * time.Second, 2 * time.Minute},
+			Clock:        clock.Real(),
+			Log:          discardLogger(),
+		})
+	cctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool { return h.calls.Load() >= 1 }, 200*time.Millisecond, 5*time.Millisecond)
+	cancel()
+	w.Wait()
+
+	got, err := repo.Get(ctx, enqueued.ID)
+	require.NoError(t, err)
+	assert.Equal(t, jobs.StatusPending, got.Status)
+	assert.Equal(t, int32(1), got.Attempts)
+	require.NotNil(t, got.LastError)
+	assert.Equal(t, "transient", *got.LastError)
+}
+
+func TestWorker_PermanentErrorDeadLetters(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	enqueued, err := repo.Enqueue(ctx, jobs.TypeWebhookDeliver, json.RawMessage(`{}`), time.Now(), 5)
+	require.NoError(t, err)
+	h := &recordingHandler{typ: jobs.TypeWebhookDeliver, wantErr: jobs.Permanent(errors.New("4xx config error"))}
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{jobs.TypeWebhookDeliver: h},
+		jobs.WorkerOptions{ID: "t", PollInterval: 5 * time.Millisecond, Backoff: []time.Duration{time.Second}, Clock: clock.Real(), Log: discardLogger()})
+	cctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool {
+		got, _ := repo.Get(ctx, enqueued.ID)
+		return got != nil && got.Status == jobs.StatusDead
+	}, 300*time.Millisecond, 10*time.Millisecond)
+	cancel()
+	w.Wait()
+}
+
+func TestWorker_MaxAttemptsExhaustedDeadLetters(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	enqueued, err := repo.Enqueue(ctx, jobs.TypeWebhookDeliver, json.RawMessage(`{}`), time.Now(), 2) // max=2
+	require.NoError(t, err)
+	h := &recordingHandler{typ: jobs.TypeWebhookDeliver, wantErr: errors.New("transient")}
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{jobs.TypeWebhookDeliver: h},
+		jobs.WorkerOptions{ID: "t", PollInterval: 2 * time.Millisecond, Backoff: []time.Duration{time.Millisecond}, Clock: clock.Real(), Log: discardLogger()})
+	cctx, cancel := context.WithTimeout(ctx, 500*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool {
+		got, _ := repo.Get(ctx, enqueued.ID)
+		return got != nil && got.Status == jobs.StatusDead
+	}, 500*time.Millisecond, 10*time.Millisecond)
+	cancel()
+	w.Wait()
+}
+
+func TestWorker_HandlerPanicTreatedAsRetry(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	enqueued, err := repo.Enqueue(ctx, jobs.TypeWebhookDeliver, json.RawMessage(`{}`), time.Now(), 5)
+	require.NoError(t, err)
+	h := &recordingHandler{typ: jobs.TypeWebhookDeliver, beforeRet: func() { panic("kaboom") }}
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{jobs.TypeWebhookDeliver: h},
+		jobs.WorkerOptions{ID: "t", PollInterval: 5 * time.Millisecond, Backoff: []time.Duration{time.Second}, Clock: clock.Real(), Log: discardLogger()})
+	cctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool {
+		got, _ := repo.Get(ctx, enqueued.ID)
+		return got != nil && got.Status == jobs.StatusPending && got.Attempts >= 1
+	}, 300*time.Millisecond, 10*time.Millisecond)
+	cancel()
+	w.Wait()
+}
+
+func TestWorker_UnknownTypeDeadLetters(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+	repo, _ := setupRepo(t)
+	enqueued, err := repo.Enqueue(ctx, "unknown.type", json.RawMessage(`{}`), time.Now(), 5)
+	require.NoError(t, err)
+	w := jobs.NewWorker(repo, map[jobs.JobType]jobs.Handler{},
+		jobs.WorkerOptions{ID: "t", PollInterval: 5 * time.Millisecond, Backoff: []time.Duration{time.Second}, Clock: clock.Real(), Log: discardLogger()})
+	cctx, cancel := context.WithTimeout(ctx, 200*time.Millisecond)
+	defer cancel()
+	go w.Run(cctx)
+	require.Eventually(t, func() bool {
+		got, _ := repo.Get(ctx, enqueued.ID)
+		return got != nil && got.Status == jobs.StatusDead
+	}, 200*time.Millisecond, 10*time.Millisecond)
+	cancel()
+	w.Wait()
+}