fix(kafka): retain failed message persistence

server/kafka_runtime_batch_test.go

@@ -3,6 +3,7 @@ package main
 import (
 	"context"
 	"errors"
+	"strings"
 	"testing"
 	"time"
 
@@ -47,6 +48,33 @@ func TestConsumeKafkaMessageBatchRetriesHandlerAndCommit(t *testing.T) {
 	}
 }
 
+func TestConsumeKafkaMessageBatchDoesNotCommitWhileHandlerFails(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	handlerCalls := 0
+	commitCalls := 0
+
+	err := consumeKafkaMessageBatch(ctx, []kafka.Message{{Offset: 1}},
+		func(context.Context, []kafka.Message) error {
+			handlerCalls++
+			return errors.New("database unavailable")
+		},
+		func(context.Context, ...kafka.Message) error {
+			commitCalls++
+			return nil
+		},
+		func(ctx context.Context, _ time.Duration) error {
+			cancel()
+			return ctx.Err()
+		})
+
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("consumeKafkaMessageBatch() error = %v, want context canceled", err)
+	}
+	if handlerCalls != 1 || commitCalls != 0 {
+		t.Fatalf("calls handler=%d commit=%d, want 1/0", handlerCalls, commitCalls)
+	}
+}
+
 func TestConsumeKafkaMalformedMessageRetriesCommit(t *testing.T) {
 	commitCalls := 0
 	waitCalls := 0
@@ -76,10 +104,9 @@ func TestConsumeKafkaMalformedMessageRetriesCommit(t *testing.T) {
 	}
 }
 
-func TestRunKafkaMessageSingleRetriesThenDeadLetters(t *testing.T) {
+func TestRunKafkaMessageSingleRetriesReturnsErrorWithoutTerminalCommit(t *testing.T) {
 	operationCalls := 0
 	markFailedCalls := 0
-	markDeadCalls := 0
 	waitCalls := 0
 
 	err := runKafkaMessageSingleRetries(context.Background(), 3,
@@ -91,31 +118,23 @@ func TestRunKafkaMessageSingleRetriesThenDeadLetters(t *testing.T) {
 			markFailedCalls++
 			return &types.KafkaConsumeRecord{RetryCount: markFailedCalls}, nil
 		},
-		func(record *types.KafkaConsumeRecord, err error) error {
-			markDeadCalls++
-			if record == nil || record.RetryCount != 4 || err == nil {
-				t.Fatalf("unexpected dead-letter input: record=%+v err=%v", record, err)
-			}
-			return nil
-		},
 		func(context.Context, time.Duration) error {
 			waitCalls++
 			return nil
 		})
 
-	if err != nil {
-		t.Fatalf("runKafkaMessageSingleRetries() error = %v", err)
+	if err == nil || !strings.Contains(err.Error(), "message persistence retries exhausted") {
+		t.Fatalf("runKafkaMessageSingleRetries() error = %v, want retry exhaustion", err)
 	}
-	if operationCalls != 4 || markFailedCalls != 4 || waitCalls != 3 || markDeadCalls != 1 {
-		t.Fatalf("calls operation=%d failed=%d wait=%d dead=%d, want 4/4/3/1",
-			operationCalls, markFailedCalls, waitCalls, markDeadCalls)
+	if operationCalls != 4 || markFailedCalls != 4 || waitCalls != 3 {
+		t.Fatalf("calls operation=%d failed=%d wait=%d, want 4/4/3",
+			operationCalls, markFailedCalls, waitCalls)
 	}
 }
 
 func TestRunKafkaMessageSingleRetriesStopsOnSuccess(t *testing.T) {
 	operationCalls := 0
 	markFailedCalls := 0
-	markDeadCalls := 0
 	err := runKafkaMessageSingleRetries(context.Background(), 3,
 		func() error {
 			operationCalls++
@@ -128,16 +147,12 @@ func TestRunKafkaMessageSingleRetriesStopsOnSuccess(t *testing.T) {
 			markFailedCalls++
 			return &types.KafkaConsumeRecord{RetryCount: markFailedCalls}, nil
 		},
-		func(*types.KafkaConsumeRecord, error) error {
-			markDeadCalls++
-			return nil
-		},
 		func(context.Context, time.Duration) error { return nil })
 
 	if err != nil {
 		t.Fatalf("runKafkaMessageSingleRetries() error = %v", err)
 	}
-	if operationCalls != 3 || markFailedCalls != 2 || markDeadCalls != 0 {
-		t.Fatalf("calls operation=%d failed=%d dead=%d, want 3/2/0", operationCalls, markFailedCalls, markDeadCalls)
+	if operationCalls != 3 || markFailedCalls != 2 {
+		t.Fatalf("calls operation=%d failed=%d, want 3/2", operationCalls, markFailedCalls)
 	}
 }