feat: expirar jobs stuck com status timeout

Adiciona mecanismo de expiração periódica para jobs que ficam presos
nos status received ou processing além de um threshold configurável.

- models: novo status StatusTimeout
- store: migration v2 com índice em jobs(status); método ExpireStuck
  que faz UPDATE atômico para os status alvo anteriores ao threshold
- expirer: novo pacote com goroutine periódica que chama ExpireStuck
  e loga WARN quando algum job é expirado
- config: campos JobTimeoutMinutes (default 120) e
  ExpirerIntervalMinutes (default 10), lidos via env vars
  JOB_TIMEOUT_MINUTES e EXPIRER_INTERVAL_MINUTES
- main: instancia e inicia o Expirer no startup
- dashboard: card Timeout com contagem; badge neutral na tabela de jobs

cmd/server/main.go

@@ -12,6 +12,7 @@ import (
 	"time"
 
 	"coda/internal/config"
+	"coda/internal/expirer"
 	"coda/internal/jobs"
 	"coda/internal/lidarr"
 	"coda/internal/logger"
@@ -74,6 +75,13 @@ func main() {
 	}
 	pool.Start(ctx)
 
+	exp := expirer.New(
+		st.Jobs,
+		time.Duration(envCfg.JobTimeoutMinutes)*time.Minute,
+		time.Duration(envCfg.ExpirerIntervalMinutes)*time.Minute,
+	)
+	exp.Start(ctx)
+
 	r := chi.NewRouter()
 	r.Use(middleware.RequestID)
 	r.Use(middleware.Recoverer)

internal/config/config.go

@@ -68,6 +68,8 @@ type Config struct {
 	NtfyURL               string
 	NtfyTopic             string
 	NtfyToken             string
+	JobTimeoutMinutes     int
+	ExpirerIntervalMinutes int
 }
 
 func Load() (*Config, error) {
@@ -102,6 +104,8 @@ func buildConfig(overrides map[string]string) *Config {
 		NtfyURL:                mergeString(overrides, "NTFY_URL", ""),
 		NtfyTopic:              mergeString(overrides, "NTFY_TOPIC", ""),
 		NtfyToken:              mergeString(overrides, "NTFY_TOKEN", ""),
+		JobTimeoutMinutes:      mergeInt(overrides, "JOB_TIMEOUT_MINUTES", 120),
+		ExpirerIntervalMinutes: mergeInt(overrides, "EXPIRER_INTERVAL_MINUTES", 10),
 	}
 }
 

internal/expirer/expirer.go

@@ -0,0 +1,57 @@
+package expirer
+
+import (
+	"context"
+	"log/slog"
+	"time"
+
+	"coda/internal/models"
+)
+
+type JobExpirer interface {
+	ExpireStuck(ctx context.Context, statuses []models.JobStatus, olderThan time.Duration) (int, error)
+}
+
+type Expirer struct {
+	jobs     JobExpirer
+	timeout  time.Duration
+	interval time.Duration
+}
+
+func New(jobs JobExpirer, timeout, interval time.Duration) *Expirer {
+	return &Expirer{
+		jobs:     jobs,
+		timeout:  timeout,
+		interval: interval,
+	}
+}
+
+func (e *Expirer) Start(ctx context.Context) {
+	go e.run(ctx)
+}
+
+func (e *Expirer) run(ctx context.Context) {
+	ticker := time.NewTicker(e.interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+			e.expire(ctx)
+		}
+	}
+}
+
+func (e *Expirer) expire(ctx context.Context) {
+	statuses := []models.JobStatus{models.StatusReceived, models.StatusProcessing}
+	n, err := e.jobs.ExpireStuck(ctx, statuses, e.timeout)
+	if err != nil {
+		slog.Error("expirer: failed to expire stuck jobs", "error", err)
+		return
+	}
+	if n > 0 {
+		slog.Warn("expirer: expired stuck jobs", "count", n, "timeout", e.timeout)
+	}
+}

internal/models/models.go

@@ -9,6 +9,7 @@ const (
 	StatusProcessing JobStatus = "processing"
 	StatusDone       JobStatus = "done"
 	StatusError      JobStatus = "error"
+	StatusTimeout    JobStatus = "timeout"
 )
 
 type JobSource string

internal/ntfy/client.go

@@ -20,12 +20,7 @@ const (
 	maxResponseBytes = 4096
 )
 
-type Client struct {
-	cfg *config.Service
-	hc  *http.Client
-}
-
-type message struct {
+type payload struct {
 	Topic    string   `json:"topic"`
 	Title    string   `json:"title"`
 	Message  string   `json:"message"`
@@ -33,6 +28,11 @@ type message struct {
 	Tags     []string `json:"tags"`
 }
 
+type Client struct {
+	cfg *config.Service
+	hc  *http.Client
+}
+
 func New(cfg *config.Service) *Client {
 	return &Client{
 		cfg: cfg,
@@ -62,13 +62,14 @@ func (c *Client) Notify(ctx context.Context, job *models.Job) error {
 		return nil
 	}
 
-	msg := buildMessage(job, cfg.NtfyTopic)
-	body, err := json.Marshal(msg)
+	p := buildPayload(cfg.NtfyTopic, job)
+
+	body, err := json.Marshal(p)
 	if err != nil {
-		return fmt.Errorf("marshal notification: %w", err)
+		return fmt.Errorf("marshal ntfy payload: %w", err)
 	}
 
-	endpoint, err := rootURL(cfg.NtfyURL)
+	endpoint, err := publishURL(cfg.NtfyURL)
 	if err != nil {
 		return err
 	}
@@ -97,9 +98,9 @@ func (c *Client) Notify(ctx context.Context, job *models.Job) error {
 	return fmt.Errorf("ntfy returned %d: %s", resp.StatusCode, string(respBody))
 }
 
-func buildMessage(job *models.Job, topic string) message {
+func buildPayload(topic string, job *models.Job) payload {
 	if job.Status == models.StatusError {
-		return message{
+		return payload{
 			Topic:    topic,
 			Title:    fmt.Sprintf("Coda: %s - %s failed", job.Artist, job.Album),
 			Message:  job.ErrorMsg,
@@ -107,8 +108,7 @@ func buildMessage(job *models.Job, topic string) message {
 			Tags:     []string{"x"},
 		}
 	}
-
-	return message{
+	return payload{
 		Topic:    topic,
 		Title:    fmt.Sprintf("Coda: %s - %s completed", job.Artist, job.Album),
 		Message:  "Conversion completed successfully",
@@ -117,7 +117,7 @@ func buildMessage(job *models.Job, topic string) message {
 	}
 }
 
-func rootURL(baseURL string) (string, error) {
+func publishURL(baseURL string) (string, error) {
 	u, err := url.Parse(baseURL)
 	if err != nil {
 		return "", fmt.Errorf("parse ntfy URL: %w", err)

internal/ntfy/client_test.go

@@ -3,6 +3,7 @@ package ntfy
 import (
 	"context"
 	"encoding/json"
+	"io"
 	"net/http"
 	"net/http/httptest"
 	"sync/atomic"
@@ -46,11 +47,25 @@ func newClientWithOverrides(t *testing.T, overrides map[string]string) *Client {
 	return New(svc)
 }
 
+func readPayload(t *testing.T, r *http.Request) payload {
+	t.Helper()
+	b, err := io.ReadAll(r.Body)
+	if err != nil {
+		t.Fatalf("read body: %v", err)
+	}
+	var p payload
+	if err := json.Unmarshal(b, &p); err != nil {
+		t.Fatalf("unmarshal payload: %v (body: %s)", err, string(b))
+	}
+	return p
+}
+
 func TestNotifySendsDoneMessage(t *testing.T) {
 	var (
 		gotMethod      string
 		gotPath        string
-		gotBody   message
+		gotContentType string
+		gotPayload     payload
 		wasCalled      atomic.Bool
 	)
 
@@ -58,9 +73,8 @@ func TestNotifySendsDoneMessage(t *testing.T) {
 		wasCalled.Store(true)
 		gotMethod = r.Method
 		gotPath = r.URL.Path
-		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
-			t.Errorf("decode body: %v", err)
-		}
+		gotContentType = r.Header.Get("Content-Type")
+		gotPayload = readPayload(t, r)
 		w.WriteHeader(http.StatusOK)
 	}))
 	defer srv.Close()
@@ -91,30 +105,31 @@ func TestNotifySendsDoneMessage(t *testing.T) {
 	if gotPath != "/" {
 		t.Errorf("path = %q, want /", gotPath)
 	}
-	if gotBody.Topic != "coda" {
-		t.Errorf("topic = %q, want coda", gotBody.Topic)
+	if gotContentType != "application/json" {
+		t.Errorf("Content-Type = %q, want application/json", gotContentType)
 	}
-	if gotBody.Title != "Coda: Artist - Album completed" {
-		t.Errorf("title = %q, want artist and album completed title", gotBody.Title)
+	if gotPayload.Topic != "coda" {
+		t.Errorf("topic = %q, want coda", gotPayload.Topic)
 	}
-	if gotBody.Message != "Conversion completed successfully" {
-		t.Errorf("message = %q, want success message", gotBody.Message)
+	if gotPayload.Title != "Coda: Artist - Album completed" {
+		t.Errorf("title = %q, want completed title", gotPayload.Title)
 	}
-	if gotBody.Priority != "default" {
-		t.Errorf("priority = %q, want default", gotBody.Priority)
+	if gotPayload.Message != "Conversion completed successfully" {
+		t.Errorf("message = %q, want success message", gotPayload.Message)
 	}
-	if len(gotBody.Tags) != 1 || gotBody.Tags[0] != "white_check_mark" {
-		t.Errorf("tags = %v, want [white_check_mark]", gotBody.Tags)
+	if gotPayload.Priority != "default" {
+		t.Errorf("priority = %q, want default", gotPayload.Priority)
+	}
+	if len(gotPayload.Tags) != 1 || gotPayload.Tags[0] != "white_check_mark" {
+		t.Errorf("tags = %v, want [white_check_mark]", gotPayload.Tags)
 	}
 }
 
 func TestNotifySendsErrorMessage(t *testing.T) {
-	var gotBody message
+	var gotPayload payload
 
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
-			t.Errorf("decode body: %v", err)
-		}
+		gotPayload = readPayload(t, r)
 		w.WriteHeader(http.StatusOK)
 	}))
 	defer srv.Close()
@@ -137,17 +152,50 @@ func TestNotifySendsErrorMessage(t *testing.T) {
 		t.Fatalf("Notify: %v", err)
 	}
 
-	if gotBody.Title != "Coda: Artist - Album failed" {
-		t.Errorf("title = %q, want artist and album failed title", gotBody.Title)
+	if gotPayload.Title != "Coda: Artist - Album failed" {
+		t.Errorf("title = %q, want failed title", gotPayload.Title)
 	}
-	if gotBody.Message != "validation failed" {
-		t.Errorf("message = %q, want error message", gotBody.Message)
+	if gotPayload.Message != "validation failed" {
+		t.Errorf("message = %q, want error message", gotPayload.Message)
 	}
-	if gotBody.Priority != "high" {
-		t.Errorf("priority = %q, want high", gotBody.Priority)
+	if gotPayload.Priority != "high" {
+		t.Errorf("priority = %q, want high", gotPayload.Priority)
 	}
-	if len(gotBody.Tags) != 1 || gotBody.Tags[0] != "x" {
-		t.Errorf("tags = %v, want [x]", gotBody.Tags)
+	if len(gotPayload.Tags) != 1 || gotPayload.Tags[0] != "x" {
+		t.Errorf("tags = %v, want [x]", gotPayload.Tags)
+	}
+}
+
+func TestNotifyErrorMsgWithJSONContent(t *testing.T) {
+	var gotPayload payload
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPayload = readPayload(t, r)
+		w.WriteHeader(http.StatusOK)
+	}))
+	defer srv.Close()
+
+	c := newClientWithOverrides(t, map[string]string{
+		"DRY_RUN":      "false",
+		"NTFY_ENABLED": "true",
+		"NTFY_URL":     srv.URL,
+		"NTFY_TOPIC":   "coda",
+	})
+
+	jsonErr := `{"error": "ffmpeg exit code 1", "details": "no such file"}`
+	err := c.Notify(context.Background(), &models.Job{
+		ID:       "job-1",
+		Artist:   "Artist",
+		Album:    "Album",
+		Status:   models.StatusError,
+		ErrorMsg: jsonErr,
+	})
+	if err != nil {
+		t.Fatalf("Notify: %v", err)
+	}
+
+	if gotPayload.Message != jsonErr {
+		t.Errorf("message = %q, want raw JSON string as message", gotPayload.Message)
 	}
 }
 

internal/store/jobs.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"database/sql"
 	"fmt"
+	"strings"
 	"time"
 
 	"coda/internal/models"
@@ -145,3 +146,35 @@ func (r *JobRepository) Finish(ctx context.Context, id string, status models.Job
 	}
 	return nil
 }
+
+func (r *JobRepository) ExpireStuck(ctx context.Context, statuses []models.JobStatus, olderThan time.Duration) (int, error) {
+	if len(statuses) == 0 {
+		return 0, nil
+	}
+
+	placeholders := make([]string, len(statuses))
+	args := make([]any, 0, len(statuses)+1)
+	for i, status := range statuses {
+		placeholders[i] = "?"
+		args = append(args, status)
+	}
+	args = append(args, time.Now().UTC().Add(-olderThan))
+
+	query := fmt.Sprintf(
+		`UPDATE jobs SET status = ?, finished_at = ?, error_msg = ? WHERE status IN (%s) AND created_at < ?`,
+		strings.Join(placeholders, ","),
+	)
+
+	finalArgs := []any{models.StatusTimeout, time.Now().UTC(), "job timed out"}
+	finalArgs = append(finalArgs, args...)
+
+	result, err := r.db.ExecContext(ctx, query, finalArgs...)
+	if err != nil {
+		return 0, fmt.Errorf("expire stuck jobs: %w", err)
+	}
+	rows, err := result.RowsAffected()
+	if err != nil {
+		return 0, fmt.Errorf("rows affected: %w", err)
+	}
+	return int(rows), nil
+}

internal/store/migrations.go

@@ -59,6 +59,12 @@ func migrate(ctx context.Context, db *sql.DB) error {
 				)`,
 			},
 		},
+		{
+			version: 2,
+			sql: []string{
+				`CREATE INDEX IF NOT EXISTS idx_jobs_status ON jobs(status)`,
+			},
+		},
 	}
 
 	for _, m := range migrations {

internal/web/dashboard.go

@@ -23,6 +23,7 @@ type dashboardData struct {
 	ProcessingCount   int
 	DoneCount         int
 	ErrorCount        int
+	TimeoutCount      int
 }
 
 func (h *Handler) Dashboard(w http.ResponseWriter, r *http.Request) {
@@ -45,6 +46,7 @@ func (h *Handler) Dashboard(w http.ResponseWriter, r *http.Request) {
 	dd.ProcessingCount = h.jobCount(ctx, models.StatusProcessing)
 	dd.DoneCount = h.jobCount(ctx, models.StatusDone)
 	dd.ErrorCount = h.jobCount(ctx, models.StatusError)
+	dd.TimeoutCount = h.jobCount(ctx, models.StatusTimeout)
 
 	h.render.Render(w, "dashboard.html", dd)
 }
@@ -64,6 +66,7 @@ func (h *Handler) Status(w http.ResponseWriter, r *http.Request) {
 		ProcessingCount:   h.jobCount(ctx, models.StatusProcessing),
 		DoneCount:         h.jobCount(ctx, models.StatusDone),
 		ErrorCount:        h.jobCount(ctx, models.StatusError),
+		TimeoutCount:      h.jobCount(ctx, models.StatusTimeout),
 	}
 
 	h.render.Render(w, "status.html", dd)

internal/web/library.go

@@ -1,6 +1,7 @@
 package web
 
 import (
+	"context"
 	"encoding/json"
 	"log/slog"
 	"net/http"
@@ -107,7 +108,7 @@ func (h *Handler) LibraryConvert(w http.ResponseWriter, r *http.Request) {
 		album := filepath.Base(absPath)
 		artist := filepath.Base(filepath.Dir(absPath))
 
-		id, err := h.enqueuer.Enqueue(r.Context(), jobs.EnqueueInput{
+		id, err := h.enqueuer.Enqueue(context.Background(), jobs.EnqueueInput{
 			Artist: artist,
 			Album:  album,
 			Path:   absPath,

internal/web/render.go

@@ -81,6 +81,8 @@ func statusBadge(level string) template.HTML {
 		return template.HTML(`<span class="badge badge-error">` + level + `</span>`)
 	case "received":
 		return template.HTML(`<span class="badge badge-info">` + level + `</span>`)
+	case "timeout":
+		return template.HTML(`<span class="badge badge-neutral">` + level + `</span>`)
 	default:
 		return template.HTML(level)
 	}

internal/webhook/handler.go

@@ -1,6 +1,7 @@
 package webhook
 
 import (
+	"context"
 	"encoding/json"
 	"fmt"
 	"log/slog"
@@ -89,7 +90,7 @@ func (h *Handler) Handle(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	id, err := h.enqueuer.Enqueue(r.Context(), jobs.EnqueueInput{
+	id, err := h.enqueuer.Enqueue(context.Background(), jobs.EnqueueInput{
 		Artist: payload.Artist.Name,
 		Album:  payload.Album.Title,
 		Path:   resolved,

internal/worker/worker.go

@@ -120,6 +120,8 @@ func (p *Pool) process(ctx context.Context, id string) {
 		return
 	}
 
+	logger = slog.With("job_id", job.ID, "artist", job.Artist, "album", job.Album)
+
 	flacFiles, err := filepath.Glob(filepath.Join(job.Path, "*.flac"))
 	if err != nil {
 		p.finishJob(ctx, logger, job, models.StatusError, time.Since(startedAt), fmt.Sprintf("glob flac files: %v", err))
@@ -137,8 +139,11 @@ func (p *Pool) process(ctx context.Context, id string) {
 		return
 	}
 
+	logger.Info("starting job", "path", job.Path, "flac_count", len(flacFiles))
+
 	var mp3Paths []string
 	for _, flac := range flacFiles {
+		logger.Info("converting file", "file", filepath.Base(flac))
 		mp3, err := p.transcoder.ConvertFile(ctx, flac)
 		if err != nil {
 			p.finishJob(ctx, logger, job, models.StatusError, time.Since(startedAt), fmt.Sprintf("convert %s: %v", filepath.Base(flac), err))
@@ -165,13 +170,6 @@ func (p *Pool) process(ctx context.Context, id string) {
 				logger.Warn("rescanner failed", "error", err)
 			}
 		}
-
-		logger.Info("job completed",
-			"artist", job.Artist,
-			"album", job.Album,
-			"files", len(flacFiles),
-			"duration_ms", time.Since(startedAt).Milliseconds(),
-		)
 	}
 
 	p.finishJob(ctx, logger, job, models.StatusDone, time.Since(startedAt), "")
@@ -193,4 +191,10 @@ func (p *Pool) finishJob(ctx context.Context, logger *slog.Logger, job *models.J
 	if err := p.store.Jobs.Finish(ctx, job.ID, status, durationMs, errMsg); err != nil {
 		logger.Error("failed to finish job", "error", err)
 	}
+
+	if status == models.StatusError {
+		logger.Error("job failed", "error_msg", errMsg, "duration_ms", durationMs)
+	} else {
+		logger.Info("job completed", "duration_ms", durationMs)
+	}
 }

web/templates/dashboard.html

@@ -31,7 +31,7 @@
         </div>
     </div>
 
-    <div class="grid grid-cols-2 md:grid-cols-4 gap-4">
+    <div class="grid grid-cols-2 md:grid-cols-5 gap-4">
         <div class="card bg-base-100 shadow">
             <div class="card-body">
                 <h2 class="card-title">Received</h2>
@@ -56,6 +56,12 @@
                 <p class="text-4xl font-bold text-error">{{ .ErrorCount }}</p>
             </div>
         </div>
+        <div class="card bg-base-100 shadow">
+            <div class="card-body">
+                <h2 class="card-title">Timeout</h2>
+                <p class="text-4xl font-bold text-neutral">{{ .TimeoutCount }}</p>
+            </div>
+        </div>
     </div>
 
     <div class="card bg-base-100 shadow">