From 5d9460877de5bfabae37503f64a3221db00229e5 Mon Sep 17 00:00:00 2001
From: Alex Guerrieri <hoi@klaidliadon.dev>
Date: Mon, 19 Oct 2026 12:56:26 +0200
Subject: [PATCH] feat: log every failed attempt that will be retried

- only a parked task was logged, so retries were invisible until the
  budget ran out

Claude-Session: https://claude.ai/code/session_01JeE7nT7xaWppRm9ngjdD6B
---
 executor.go      |  4 ++++
 executor_test.go | 20 ++++++++++++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/executor.go b/executor.go
index 8d33867..a07e1f3 100644
--- a/executor.go
+++ b/executor.go
@@ -205,6 +205,10 @@ func (m *Manager) executePlan(ctx context.Context, claimed *store.ClaimedRun) St
 			return StatusStuck
 		}
 
+		taskLog.WarnContext(ctx, "workflow: task attempt failed",
+			slog.Int("attempt", attempt), slog.Int("budget", budget),
+			slog.Duration("backoff", backoff), slog.Any("error", err))
+
 		if backoff <= m.cfg.MaxInProcessBackoff {
 			persisted, perr := m.db.PersistTaskFailure(ctx, runID, seq, failure, backoff, false, claimID)
 			if perr != nil {
diff --git a/executor_test.go b/executor_test.go
index d297af7..04cb4c0 100644
--- a/executor_test.go
+++ b/executor_test.go
@@ -177,6 +177,16 @@ func (h *recordingHandler) messages() []string {
 	return slices.Clone(h.msgs)
 }
 
+func countMessage(h *recordingHandler, msg string) int {
+	n := 0
+	for _, m := range h.messages() {
+		if m == msg {
+			n++
+		}
+	}
+	return n
+}
+
 func TestNextBackoff(t *testing.T) {
 	cfg := Config{BackoffBase: time.Second, BackoffMax: time.Minute}
 	schedule := []time.Duration{time.Second, 5 * time.Second, 30 * time.Second}
@@ -673,11 +683,14 @@ func TestExecutor(t *testing.T) {
 				return map[string]string{"v": "ok"}, nil
 			}, Backoff(10*time.Millisecond, 50*time.Millisecond), MaxAttempts(5)),
 		)))
-		mgr := newManager(t, db, reg, func(c *Config) { c.MaxInProcessBackoff = time.Second })
+		rec := new(recordingHandler)
+		mgr, err := NewManager(t.Context(), db.DB, reg, Config{MaxInProcessBackoff: time.Second}, slog.New(rec))
+		require.NoError(t, err)
 
 		plan := seedPlan(t, db, ctx, mgr, "test.retry", testInput{Seed: "x"})
 		require.Equal(t, StatusSucceeded, mgr.executePlan(ctx, claim(t, db, ctx, plan.ID)))
 		require.EqualValues(t, 3, calls.Load())
+		require.Equal(t, 2, countMessage(rec, "workflow: task attempt failed"))
 
 		got, err := db.GetRun(ctx, plan.ID)
 		require.NoError(t, err)
@@ -695,11 +708,14 @@ func TestExecutor(t *testing.T) {
 				return nil, errors.New("transient")
 			}, Backoff(20*time.Second, time.Minute), MaxAttempts(5)),
 		)))
-		mgr := newManager(t, db, reg)
+		rec := new(recordingHandler)
+		mgr, err := NewManager(t.Context(), db.DB, reg, Config{}, slog.New(rec))
+		require.NoError(t, err)
 
 		plan := seedPlan(t, db, ctx, mgr, "test.longbackoff", testInput{Seed: "x"})
 		before := time.Now()
 		require.Equal(t, StatusRunning, mgr.executePlan(ctx, claim(t, db, ctx, plan.ID)))
+		require.Equal(t, 1, countMessage(rec, "workflow: task attempt failed"))
 
 		got, err := db.GetRun(ctx, plan.ID)
 		require.NoError(t, err)