Exercise - Build and publish cross-platform DeepStream container images

10 minutes

The DeepStream 6.0 container-builder package can build cross-platform container images. You'll need to use this capability to build a container that's compatible with the ARM64 CPU on your target NVIDIA embedded device.

Complete the following steps on an x86-based host machine that has DeepStream 6.0 Graph Composer with Docker installed. The steps assume that you completed the steps described in Set up and configure an NVIDIA DeepStream development environment and Introduction to NVIDIA DeepStream Graph Composer with Azure.

To begin, you must install the qemu-user-static and binfmt-support packages. The packages are dependencies that are needed for cross-platform support in Docker. To perform this step, run the following commands in a terminal on the host:
```
sudo apt install qemu-user-static binfmt-support
```
When installation is finished, run the following commands on the host to configure cross-platform support in Docker:
```
sudo docker run --rm --privileged multiarch/qemu-user-static --reset --persistent yes --credential yes 
```

Next, you'll create a container-builder configuration that will produce an ARM64 based container to support the deepstream-test4 graph. To begin, go to /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test4 and open the ds_test4_container_builder_jetson.yaml file with:

    cd /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test4
    sudo vi ds_test4_container_builder_jetson.yaml

The contents of the unmodified ds_test4_container_builder_jetson.yaml file are shown below:

    %YAML 1.2
    ################################################################################
    # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
    #
    # NVIDIA Corporation and its licensors retain all intellectual property
    # and proprietary rights in and to this software, related documentation
    # and any modifications thereto.  Any use, reproduction, disclosure or
    # distribution of this software and related documentation without an express
    # license agreement from NVIDIA Corporation is strictly prohibited.
    #
    ################################################################################
    ---
    # final stage
    unique_stage: final_image
    base_image: auto
    stage_model: clean_stage

    local_copy_files:
    - src: "/opt/nvidia/graph-composer/graph_composer-1.0.0_arm64.deb"
      dst: "/tmp/"
    - src: "parameters.yaml"
      dst: "/workspace/deepstream-test4/" # ends with '/' is folder

    custom_runs:
     - "apt install -y /tmp/graph_composer-1.0.0_arm64.deb && rm -rf /tmp/graph-composer-1.0.0_arm64.deb"

    work_folder: /workspace/deepstream-test4/

    env_list:
      PATH: "/opt/nvidia/graph-composer/:$PATH"
      LD_LIBRARY_PATH: "/opt/nvidia/graph-composer/:$LD_LIBRARY_PATH"
      DISPLAY: ":0"

    # Setup entrypoint
    entrypoint:
    - "gxe"
    - "--manifest"
    - "deepstream-test4_manifest.yaml"
    - "--app"
    - "deepstream-test4.yaml,parameters.yaml"
    ---
    # Container Builder Config
    container_builder: main # required
    graph:
      graph_files: [ deepstream-test4.yaml ]
      graph_dst: /workspace/deepstream-test4/
      manifest_dst: /workspace/deepstream-test4/deepstream-test4_manifest.yaml
      ext_install_root: /workspace/deepstream-test4/
    docker_build:
        image_name: <docker-container-registry>:deepstream-test4-jetson

Note

Here are some tips for using the vi editor when you edit the ds_test4_container_builder_jetson.yaml file:

To enter insert mode, select i. Then, you can make changes.
To exit insert mode and return to normal mode, select Esc.
To save the file and quit, select x, and then select Enter.
To save the file, select w, and then select Enter.
To quit vi, enter Quit, and then select Enter.

Take note of the very last line of this file as it will require a modification:

image_name: <docker-container-registry>:deepstream-test4-jetson

Update this section to look like the following:

image_name: deepstream-test4-jetson

Note

If you do not modify this section, attempts to build the cross-platform image in later steps will produce the following error: ConnectionResetError: [Errno 104] Connection reset by peer

Take note that this specification uses a base image from nvcr.io and copies in the deepstream-test4.yaml and parameters.yaml files. The image entrypoint starts the deepstream-test4 graph and overrides it with the values present in parameters.yaml. For this reason, we also need to update this file before we actually build the container.

Return to the directory of the deepstream-test4 reference graphs and edit the parameters.yaml file:

cd /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test4
sudo vi parameters.yaml

Change the value of msg-broker-proto-lib to /opt/nvidia/deepstream/deepstream/lib/libnvds_azure_edge_proto.so.

Also, change the value of msg-conv-payload-type to 1. This parameter controls how the resulting message output is formatted. If you don't change this parameter, the output uses a hard-coded format that's provided in the NVIDIA samples. The format might not represent your object detection classes if you update the model used in NvDsInferVideo.

The final contents of parameters.yaml should look like the following example:

################################################################################
# Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
#
# NVIDIA Corporation and its licensors retain all intellectual property
# and proprietary rights in and to this software, related documentation
# and any modifications thereto. Any use, reproduction, disclosure or
# distribution of this software and related documentation without an express
# license agreement from NVIDIA Corporation is strictly prohibited.
#
################################################################################

components:
- name: nv_ds_single_src_input39..3c18
  parameters:
    uri: file:///opt/nvidia/deepstream/deepstream/samples/streams/sample_1080p_h264.mp4
name: NvDsSingleSrcInput..1b00
---
components:
- name: nv_ds_stream_mux42..36d8
  parameters:
    batch-size: 1
    height: 1080
    width: 1920
name: NvDsStreamMux..16d8
---
components:
- name: nv_ds_msg_conv_broker61..4748
  parameters:
    msg-broker-conn-str: 127.0.0.1;9092;test-topic
    msg-broker-proto-lib: /opt/nvidia/deepstream/deepstream/lib/libnvds_azure_edge_proto.so
    msg-conv-config: /opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/deepstream-test4/dstest4_msgconv_config.txt
    msg-conv-payload-type: 1
    topic: test-topic
    name: NvDsMsgConvBroker..bb00

Build the cross-platform image on the host machine by running the following command in a terminal:
```
cd /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test4
sudo container_builder -c ds_test4_container_builder_jetson.yaml -d /opt/nvidia/graph-composer/config/target_aarch64.yaml
```
After the code runs successfully, output that contains the message Successfully tagged deepstream-test4-jetson:latest appears.
Now, you're ready to publish this image to the container registry you created in the previous module. Run the following code in a terminal on the host to tag the image with a name that follows the format you used earlier. For <Login Server>, use the URL of your container registry.
```
sudo docker tag deepstream-test4-jetson <Login Server>/deepstream_test4_jetson:v1
```
With your image now properly tagged, push the image to your container registry in Azure by executing the following command in a terminal on the host. For <Login Server>, use the URL of your container registry.
```
sudo docker push <Login Server>/deepstream_test4_jetson:v1
```

With your cross-platform image now published to the Azure container registry, you're ready to provision your NVIDIA embedded hardware for the IoT Edge runtime and deploy this workload as an IoT Edge module.

Exercise - Build and publish cross-platform DeepStream container images

Feedback