【DeepStream】整合出现的问题

WinyQ02026-01-24 9:57

成功运行官方示例

整合后，启动出问题，每一步初始化都是通的，但是开始分析，管道里可能有未知错误，python程序捕捉不到异常，直接出现开始分析然后服务就重启

项目 状态 说明
WebSocket 连接 ✅ 成功 客户端 172.18.0.1 已连接
RTSP 流拉取 ✅ 成功 rtsp://192.168.1.102:8554/mystream 被接收
TensorRT 引擎加载 ✅ 成功 从 /app/models/detection/person.onnx_b1_gpu0_fp16.engine 正确反序列化
nvinfer 推理插件 ✅ 初始化成功 模型加载完成，无 NVDSINFER_*_FAILED 错误
GStreamer 管道 ✅ 启动成功 日志显示 "管道启动成功"、"DeepStream分析启动成功"
DeepStream 插件检测 ✅ 可用 nvstreammux 和 nvinfer 均能创建

日志如下：

bash 复制代码 
2026-01-21 19:00:33 - src.server.websocket_server - INFO - 新连接: ('172.18.0.1', 33188)

2026-01-21 19:00:33 - src.server.connection_manager - INFO - 新连接已添加: conn_1768993233_130497904152240, 客户端: ('172.18.0.1', 33188)

2026-01-21 19:00:36 - src.server.deepstream_manager - INFO - 启动DeepStream分析: test_stream_001, URL: rtsp://192.168.1.102:8554/mystream

2026-01-21 19:00:36 - src.server.deepstream_manager - INFO - 创建管道: test_stream_001

Unknown or legacy key specified 'is-classifier' for group [property]

2026-01-21 19:00:36 - src.server.deepstream_manager - INFO - appsink回调已设置: test_stream_001

2026-01-21 19:00:36 - src.server.deepstream_manager - INFO - 启动管道: test_stream_001

0:18:10.582165560     1 0x643276f19550 WARN                 nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Warning from NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1244> [UID = 1]: Warning, OpenCV has been deprecated. Using NMS for clustering instead of cv::groupRectangles with topK = 20 and NMS Threshold = 0.5

WARNING: [TRT]: The getMaxBatchSize() function should not be used with an engine built from a network created with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag. This function will always return 1.

INFO: ../nvdsinfer/nvdsinfer_model_builder.cpp:612 [Implicit Engine Info]: layers num: 2

0   INPUT  kFLOAT images          3x640x640       

1   OUTPUT kFLOAT output0         300x6           


0:18:18.087474452     1 0x643276f19550 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2095> [UID = 1]: deserialized trt engine from :/app/models/detection/person.onnx_b1_gpu0_fp16.engine

0:18:18.264057886     1 0x643276f19550 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::generateBackendContext() <nvdsinfer_context_impl.cpp:2198> [UID = 1]: Use deserialized engine model: /app/models/detection/person.onnx_b1_gpu0_fp16.engine

0:18:18.277659566     1 0x643276f19550 INFO                 nvinfer gstnvinfer_impl.cpp:343:notifyLoadModelStatus:<nvinfer0> [UID 1]: Load new model:/app/configs/deepstream_config.txt sucessfully

2026-01-21 19:00:43 - src.server.deepstream_manager - INFO - 管道启动成功: test_stream_001

2026-01-21 19:00:43 - src.server.deepstream_manager - INFO - DeepStream分析启动成功: test_stream_001

INFO:     127.0.0.1:37238 - "GET /health HTTP/1.1" 200 OK

✓ GStreamer initialized

Starting FlexVision-Server...

Importing main module...

WARNING: PIL not available, some image processing features disabled

Trying to load plugins:

  Multistream plugin: /opt/nvidia/deepstream/deepstream/lib/gst-plugins/libnvdsgst_multistream.so

  Infer plugin: /opt/nvidia/deepstream/deepstream/lib/gst-plugins/libnvdsgst_infer.so

Plugin load results:

  Multistream plugin: <Gst.Plugin object at 0x7255ce3d5600 (GstPlugin at 0x5a633f9ed7a0)>

  Infer plugin: <Gst.Plugin object at 0x7255cb010f00 (GstPlugin at 0x5a633f9ed370)>

Element creation results:

  nvstreammux: <__gi__.GstNvStreamMux object at 0x7255c83d9080 (GstNvStreamMux at 0x5a63401044d0)>

  nvinfer: <__gi__.GstNvInfer object at 0x7255cafeef40 (GstNvInfer at 0x5a63401059e0)>

DeepStream plugins available: nvstreammux=True, nvinfer=True
1. Python 主进程崩溃（最可能，但尝试修改无果）
  • 虽然 DeepStream 管道启动成功，但你的 Python 应用（如 Uvicorn/FastAPI + GStreamer appsink 回调）可能在处理帧时出错。
  • 例如：
    • 在 appsink 回调中访问了无效内存（GstBuffer 已释放）
    • 尝试用 PIL 处理图像但未安装 → 抛出 ImportError 或 AttributeError
    • 多线程/异步操作冲突（如在非主线程调用 GTK/GStreamer 不安全 API）

💥 一旦 Python 主进程崩溃，整个容器或服务就会退出，导致 WebSocket 断开 + 服务重启。

2. GStreamer 管道错误未被捕获（很有可能）
  • 虽然管道启动成功，但后续流中断（如 RTSP 源断开、网络抖动）可能触发 error 或 eos 信号。
  • 如果你的代码没有监听 bus 消息并处理错误，可能导致未预期行为。
3. 内存或 GPU 资源耗尽（不存在）
  • DeepStream 占用较高 GPU 显存，若同时运行多个流或模型，可能触发 OOM（Out-Of-Memory）。
  • 容器被系统 kill（查看 docker logs <container> 是否有 Killed 字样）。
4. WebSocket 心跳超时或客户端主动断开（排查过不太可能）
  • 较少见，但如果你的前端长时间未发送 ping，或后端未实现 pong 响应，连接可能被关闭。
  • 但这种情况不会导致服务重启，只会断开连接。

可能缺少.engine模型，或者找不到

DeepStream 尝试加载一个 TensorRT 引擎文件 （.engine）：

/app/models/detection/person.onnx_b1_gpu0_fp16.engine

但该文件不存在。

于是它回退到尝试从原始模型重建引擎 （比如 .onnx 文件），但在目录 /app/models/detection/ 中也找不到对应的 ONNX 或其他格式的原始模型文件，导致构建失败。

bash 复制代码 
2026-01-21 15:22:07 - src.server.websocket_server - INFO - WebSocket服务器已启动，监听地址: 0.0.0.0:8765

INFO:     Uvicorn running on http://0.0.0.0:8080⁠ (Press CTRL+C to quit)

2026-01-21 15:22:19 - src.server.websocket_server - INFO - 新连接: ('172.17.0.1', 33070)

2026-01-21 15:22:19 - src.server.connection_manager - INFO - 新连接已添加: conn_1768980139_123685807818160, 客户端: ('172.17.0.1', 33070)

2026-01-21 15:22:22 - src.server.deepstream_manager - INFO - 启动DeepStream分析: test_stream_001, URL: rtsp://192.168.1.102:8554/mystream

2026-01-21 15:22:22 - src.server.deepstream_manager - INFO - 创建管道: test_stream_001

0:00:14.946489078     1 0x5eef2a9acbc0 WARN                 nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Warning from NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1244> [UID = 1]: Warning, OpenCV has been deprecated. Using NMS for clustering instead of cv::groupRectangles with topK = 20 and NMS Threshold = 0.5

0:00:21.523934791     1 0x5eef2a9acbc0 WARN                 nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Warning from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2083> [UID = 1]: deserialize engine from file :/app/models/detection/person.onnx_b1_gpu0_fp16.engine failed

WARNING: ../nvdsinfer/nvdsinfer_model_builder.cpp:1494 Deserialize engine failed because file path: /app/models/detection/person.onnx_b1_gpu0_fp16.engine open error

0:00:21.683380121     1 0x5eef2a9acbc0 WARN                 nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Warning from NvDsInferContextImpl::generateBackendContext() <nvdsinfer_context_impl.cpp:2188> [UID = 1]: deserialize backend context from engine from file :/app/models/detection/person.onnx_b1_gpu0_fp16.engine failed, try rebuild

0:00:21.686198045     1 0x5eef2a9acbc0 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:2109> [UID = 1]: Trying to create engine from model files

ERROR: ../nvdsinfer/nvdsinfer_model_builder.cpp:870 failed to build network since there is no model file matched.

ERROR: ../nvdsinfer/nvdsinfer_model_builder.cpp:809 failed to build network.

0:00:26.010287886     1 0x5eef2a9acbc0 ERROR                nvinfer gstnvinfer.cpp:676:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Error in NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:2129> [UID = 1]: build engine file failed

0:00:26.170914628     1 0x5eef2a9acbc0 ERROR                nvinfer gstnvinfer.cpp:676:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Error in NvDsInferContextImpl::generateBackendContext() <nvdsinfer_context_impl.cpp:2215> [UID = 1]: build backend context failed

0:00:26.172788284     1 0x5eef2a9acbc0 ERROR                nvinfer gstnvinfer.cpp:676:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Error in NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1352> [UID = 1]: generate backend failed, check config file settings

0:00:26.173049907     1 0x5eef2a9acbc0 WARN                 nvinfer gstnvinfer.cpp:912:gst_nvinfer_start:<nvinfer0> error: Failed to create NvDsInferContext instance

0:00:26.173075735     1 0x5eef2a9acbc0 WARN                 nvinfer gstnvinfer.cpp:912:gst_nvinfer_start:<nvinfer0> error: Config file path: /app/configs/deepstream_config.txt, NvDsInfer Error: NVDSINFER_CONFIG_FAILED

还是开始分析就重启

bash 复制代码 
发起启动分析后输出如下日志，为什么会重启
2026-01-22 12:50:53 - src.server.deepstream_manager - INFO - 启动DeepStream分析: test_stream_001, URL: rtsp://192.168.1.102:8554/mystream
2026-01-22 12:50:53 - src.server.deepstream_manager - INFO - 创建管道: test_stream_001
2026-01-22 12:50:53 - src.server.deepstream_manager - INFO - appsink回调已设置: test_stream_001
2026-01-22 12:50:53 - src.server.deepstream_manager - INFO - 启动管道: test_stream_001
0:00:11.766587517     1 0x64317d657be0 WARN                 nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Warning from NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1244> [UID = 1]: Warning, OpenCV has been deprecated. Using NMS for clustering instead of cv::groupRectangles with topK = 20 and NMS Threshold = 0.5
WARNING: [TRT]: The getMaxBatchSize() function should not be used with an engine built from a network created with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag. This function will always return 1.
INFO: ../nvdsinfer/nvdsinfer_model_builder.cpp:612 [Implicit Engine Info]: layers num: 2
0   INPUT  kFLOAT images          3x640x640       
1   OUTPUT kFLOAT output0         300x6           

0:00:18.602649702     1 0x64317d657be0 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2095> [UID = 1]: deserialized trt engine from :/app/models/detection/person.onnx_b1_gpu0_fp16.engine
0:00:18.766174530     1 0x64317d657be0 INFO                 nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<nvinfer0> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::generateBackendContext() <nvdsinfer_context_impl.cpp:2198> [UID = 1]: Use deserialized engine model: /app/models/detection/person.onnx_b1_gpu0_fp16.engine
0:00:18.774005421     1 0x64317d657be0 INFO                 nvinfer gstnvinfer_impl.cpp:343:notifyLoadModelStatus:<nvinfer0> [UID 1]: Load new model:/app/configs/deepstream_config.txt sucessfully
2026-01-22 12:51:01 - src.server.deepstream_manager - INFO - 管道启动成功: test_stream_001
2026-01-22 12:51:01 - src.server.deepstream_manager - INFO - DeepStream分析启动成功: test_stream_001
✓ GStreamer initialized
Starting FlexVision-Server...
Importing main module...
Trying to load plugins:
  Multistream plugin: /opt/nvidia/deepstream/deepstream/lib/gst-plugins/libnvdsgst_multistream.so
  Infer plugin: /opt/nvidia/deepstream/deepstream/lib/gst-plugins/libnvdsgst_infer.so
Plugin load results:
  Multistream plugin: <Gst.Plugin object at 0x76db1710e4c0 (GstPlugin at 0x5fbd6486e6a0)>
  Infer plugin: <Gst.Plugin object at 0x76db17119280 (GstPlugin at 0x5fbd6486e2c0)>
Element creation results:
  nvstreammux: <__gi__.GstNvStreamMux object at 0x76dae539c7c0 (GstNvStreamMux at 0x5fbd64f97ec0)>
  nvinfer: <__gi__.GstNvInfer object at 0x76dae5231b40 (GstNvInfer at 0x5fbd64f993d0)>
DeepStream plugins available: nvstreammux=True, nvinfer=True
Main module imported successfully
Running asyncio main...
2026-01-22 12:51:08 - src.utils.logging_utils - INFO - 日志配置完成，级别: INFO
2026-01-22 12:51:08 - src.utils.logging_utils - INFO - 日志文件: logs/inference_server.log
2026-01-22 12:51:08 - src.main - INFO - ==================================================
2026-01-22 12:51:08 - src.main - INFO - 推理服务器 V2.0 正在启动...
2026-01-22 12:51:08 - src.main - INFO - ==================================================
2026-01-22 12:51:08 - src.main - INFO - 服务器配置: {'host': '0.0.0.0', 'port': 8767, 'max_connections': 100, 'heartbeat_interval': 30, 'connection_timeout': 300}
2026-01-22 12:51:08 - src.server.connection_manager - INFO - 连接管理器初始化，最大连接数: 100
2026-01-22 12:51:08 - src.server.deepstream_manager - INFO - GStreamer initialized successfully: version (major=1, minor=20, micro=3, nano=0)
2026-01-22 12:51:08 - src.server.deepstream_manager - INFO - DeepStreamManager initialized - GStreamer: True, PyDS: True
2026-01-22 12:51:08 - src.server.websocket_server - INFO - WebSocket服务器初始化完成，配置: {'host': '0.0.0.0', 'port': 8767, 'max_connections': 100, 'heartbeat_interval': 30, 'connection_timeout': 300}
2026-01-22 12:51:08 - src.main - INFO - API服务器已启动: http://0.0.0.0:8082⁠

方案对比

方案 性能 (GPU利用率) 开发难度 灵活性 推荐场景
DeepStream ⭐⭐⭐⭐⭐ (极高) ⭐⭐⭐⭐⭐ (极难) 高 需要处理 30+ 路视频，压榨硬件极限
Savant ⭐⭐⭐⭐⭐ (极高) ⭐⭐ (简单) 高 最推荐。想要 DeepStream 的性能，但不想写 C/GStreamer
Ultralytics ⭐⭐⭐⭐ (高) ⭐ (极简) 中 只关注检测/追踪，快速落地项目
MediaPipe ⭐⭐⭐ (中) ⭐⭐ (简单) 中 需要手势、姿态等复杂交互逻辑
相关推荐
周公挚友2 小时前
centos 7.9 搭建nginx
linux·nginx·centos
坐怀不乱杯魂2 小时前
Linux 基础IO
linux·运维·服务器
小魏每天都学习2 小时前
【网络拓扑部署-网络设备-网络安全】
运维·网络
zd8451015002 小时前
CubeMX H743 lwip ETH初始化流程
网络·stm32·单片机
南棱笑笑生2 小时前
20260123让天启AIO-3576Q38开发板在天启Buildroot下读写TF卡
linux·运维·服务器·rockchip
zzh_my2 小时前
tcp 服务端（用于测试）
服务器·网络·tcp/ip
SmartRadio2 小时前
LLCC68 L型与π型匹配网络的调试方法
网络·lora·阻抗匹配·匹配·射频·llcc68
噎住佩奇2 小时前
PVC和PV等概念解释
linux·运维·服务器
lvbinemail2 小时前
配置jenkins.service
java·运维·jenkins·systemctl
热门推荐
01GitHub 镜像站点02OpenCode 入门教程：介绍 · 安装 · 配置第三方 API (如 Claude)03Claude Code Skills 实用使用手册04UV安装并设置国内源05在VSCode配置Java开发环境的保姆级教程（适配各类AI编程IDE）06BongoCat - 跨平台键盘猫动画工具07安娜的档案(Anna’s Archive) 镜像网站/国内最新可访问入口（持续更新）08AI 规范驱动开发“三剑客”深度对比：Spec-Kit、Kiro 与 OpenSpec 实战指南09Linux下V2Ray安装配置指南10Open Code教程（四）| 高级配置与集成