perf(queue): 限制大媒体任务内存并消除准入惊群

将同步与异步非文本任务的准入唤醒改为按任务和全局队首推进,批量续租等待者,避免大量请求同时争抢 advisory lock 和数据库连接。

对 Base64 请求解析、素材解码、上游媒体执行和结果物化增加分层并发限制,复用请求素材并释放重复 Gemini wire 数据;生产 API 默认入口解析并发 16,媒体物化并发 8。

新增 1000 个同步 Gemini 图像编辑请求的模拟上游压力验收,10 MiB 输入和输出下全部成功,Heap 峰值增长约 2.58 GiB,并验证共享上传哈希、单次 attempt 与本地零落盘。

验证:go test ./... -count=1;go vet ./...;gofmt -l 无输出;kubectl kustomize deploy/kubernetes/production;10 MiB Gemini Base64 千任务压力测试通过。
This commit is contained in:
2026-07-30 13:13:32 +08:00
parent fd3b6bf042
commit ea58d21d03
26 changed files with 1980 additions and 74 deletions
+54
View File
@@ -53,6 +53,60 @@ func TestGeneratedAssetDecisionUploadsInlineImageBase64(t *testing.T) {
}
}
func TestMediaResultMaterializationConcurrencyIsBounded(t *testing.T) {
service := New(config.Config{MediaMaterializationConcurrency: 1}, nil, nil)
result := map[string]any{
"data": []any{map[string]any{
"b64_json": base64.StdEncoding.EncodeToString([]byte("inline image")),
"mime_type": "image/png",
}},
}
release, err := service.acquireMediaResultSlot(context.Background(), result)
if err != nil {
t.Fatalf("acquire first media result slot: %v", err)
}
waitCtx, cancel := context.WithTimeout(context.Background(), 25*time.Millisecond)
defer cancel()
if _, err := service.acquireMediaResultSlot(waitCtx, result); err == nil {
t.Fatal("second media result materialization bypassed configured concurrency")
}
release()
release, err = service.acquireMediaResultSlot(context.Background(), result)
if err != nil {
t.Fatalf("acquire released media result slot: %v", err)
}
release()
}
func TestMediaTaskPreacquiresMaterializationSlotForInlineResultTypes(t *testing.T) {
t.Parallel()
for _, modelType := range []string{
"image_generate",
"image_edit",
"audio_generate",
"text_to_speech",
"voice_clone",
"omni",
} {
if !mediaTaskNeedsPreUpstreamMaterializationSlot(modelType) {
t.Fatalf("%s should preacquire the media materialization slot", modelType)
}
}
for _, modelType := range []string{
"text_generate",
"text_embedding",
"text_rerank",
"video_generate",
"image_to_video",
} {
if mediaTaskNeedsPreUpstreamMaterializationSlot(modelType) {
t.Fatalf("%s should not hold a media materialization slot across the upstream call", modelType)
}
}
}
func TestGeneratedAssetDecisionUploadsVectorDocumentWithoutChangingType(t *testing.T) {
item := map[string]any{
"type": "file",