Run multiple jetsons with k3s

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform (Jetson / GPU)
Jetson AGX x 2
• DeepStream Version
5.1
• JetPack Version (valid for Jetson only)
4.5.1-b17
• TensorRT Version
7.1.3-1+cuda10.2
• Issue Type( questions, new requirements, bugs)

Hi all,

I follow this tutorial: AI at the Edge with K3s and NVIDIA Jetson Nano: Object Detection and Real-Time Video Analytics and all want good until i try to deploy the 2 AGX to run simultaneous with this yaml file:

apiVersion: v1
kind: Pod
metadata:
  name: demo-pod
  labels:
    name: demo-pod
spec:
  hostNetwork: true
  nodeName: nvidia-desktop
  containers:
  - name: demo-stream
    image: nvcr.io/nvidia/deepstream-l4t:5.1-21.02-samples
    securityContext:
      privileged: true
      allowPrivilegeEscalation: true
    command:
    - sleep
    -  "150000"
    workingDir: /opt/nvidia/deepstream/deepstream-5.1
    volumeMounts:
    - mountPath: /tmp/.X11-unix/
      name: x11
    - mountPath: /dev/video0
      name: cam
  volumes:
    - name: x11
      hostPath:
        path: /tmp/.X11-unix/
    - name: cam
      hostPath:
        path: /dev/video0
---

apiVersion: v1
kind: Pod
metadata:
  name: demo-pod2
  labels:
    name: demo-pod2
spec:
  hostNetwork: true
  nodeName: nvidiaworker
  containers:
  - name: demo-stream
    image: nvcr.io/nvidia/deepstream-l4t:5.1-21.02-samples
    securityContext:
      privileged: true
      allowPrivilegeEscalation: true
    command:
    - sleep
    -  "150000"
    workingDir: /opt/nvidia/deepstream/deepstream-5.1
    volumeMounts:
    - mountPath: /tmp/.X11-unix/
      name: x11
    - mountPath: /dev/video0
      name: cam
  volumes:
    - name: x11
      hostPath:
        path: /tmp/.X11-unix/
    - name: cam
      hostPath:
        path: /dev/video0

In the master(nvidia-desktop) its work great but in the slave(nvidiaworker) i get this error :
“deepstream-app: error while loading shared libraries: /usr/lib/aarch64-linux-gnu/libnvinfer.so.7: file too short”

by running this commend:
“deepstream-app -c /opt/nvidia/deepstream/deepstream-5.1/samples/configs/deepstream-app/source4_1080p_dec_infer-resnet_tracker_sgie_tiled_display_int8.txt”

I understand I have a problem with the settings only I do not see it. If anyone knows or experiences a similar problem and knows how to solve it I would love to hear how to do it.

my goal is to run multiple deepstream-app samples simultaneity from different jetsons using k3s.

Thanks!

3

“deepstream-app: error while loading shared libraries: file too short”
please check your TensorRT installation. you can compare the two devices about this file, /usr/lib/aarch64-linux-gnu/libnvinfer.so.7

4

Thanks for the answer,
I install the two AGX in the same way and time so its must to be the same files but i still check like you ask and both files are the same size.
AGX-master:
lrwxrwxrwx 1 root root 19 Jun 6 2020 /usr/lib/aarch64-linux-gnu/libnvinfer.so.7 -> libnvinfer.so.7.1.3

AGX-worker:
lrwxrwxrwx 1 root root 19 Jun 6 2020 /usr/lib/aarch64-linux-gnu/libnvinfer.so.7 -> libnvinfer.so.7.1.3

I found the problem, i didn’t install pod network add on:
kubectl apply -f https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml

now its work.

but now i facing with new problem.

I deployed this Applications and all work great until i try to use RTSP.

I change the configmap for support RTSP and its run OK but when i open the VLC and try to run rtsp://localhost:8554/ds-test i don’t get nothing.

5

Can you try if file sink work instread of rtsp streaming? if work, you need to check your network, if port 8554 have data.

6

Thanks for the answer and sorry for the long reply.
I mange to implement this and now i study it. you can close the post and thanks for the reply!.

1 Like