Deploy with Helm#

This guide provides step-by-step instructions for deploying the MultiModal - Weld Defect Detection sample application using Helm.

Prerequisites#

System Requirements
K8s installation on single or multi node must be done as prerequisite to continue the following deployment. Note that the Kubernetes cluster is set up with kubeadm, kubectl and kubelet packages on single and multi nodes with v1.30.2. Refer to online tutorials (such as https://adamtheautomator.com/install-kubernetes-ubuntu) to setup kubernetes cluster on the web with host OS as Ubuntu 22.04.
For Helm installation, refer to helm website

Note: If Ubuntu Desktop is not installed on the target system, follow the instructions from Ubuntu to install Ubuntu desktop. The target system refers to the system where you are installing the application.

Step 1: Generate or download the Helm charts#

You can either generate or download the Helm charts.

To download the Helm charts:

Follow this procedure on the target system to install the package.

Download Helm chart with the following command:

helm pull oci://registry-1.docker.io/intel/multimodal-weld-defect-detection-sample-app --version 1.0.0-weekly

Unzip the package using the following command:

tar -xvzf multimodal-weld-defect-detection-sample-app-1.0.0-weekly.tgz

Get into the Helm directory:

cd multimodal-weld-defect-detection-sample-app

To generate the Helm charts:

cd edge-ai-suites/manufacturing-ai-suite/industrial-edge-insights-multimodal # path relative to git clone folder

make gen_helm_charts

cd helm/

Step 2: Configure and update the environment variables#

Update the following fields in the values.yaml file of the helm chart.

WORK_DIR: # Update with the absolute path to your industrial-edge-insights-multimodal directory
INFLUXDB_USERNAME:
INFLUXDB_PASSWORD:
VISUALIZER_GRAFANA_USER:
VISUALIZER_GRAFANA_PASSWORD:
HTTP_PROXY:  # example: http_proxy: http://proxy.example.com:891
HTTPS_PROXY:  # example: http_proxy: http://proxy.example.com:891
MTX_WEBRTCICESERVERS2_0_USERNAME:
MTX_WEBRTCICESERVERS2_0_PASSWORD:
HOST_IP:  # IP address of server where DL Streamer Pipeline Server is running

Step 3: Install Helm charts#

Note:

Uninstall Helm charts if already installed.

Note the helm install command fails if the above required fields are not populated as per the rules called out in the values.yaml file.

To install Helm charts, use one of the following options:

helm install multimodal-weld-defect-detection . -n multimodal-sample-app --create-namespace

Verify Installation:

Note: The deployment-coturn, deployment-fusion-analytics, deployment-ia-weld-data-simulator and deployment-telegraf pods might restart since its depended on deployment-mqtt-broker and deployment-mediamtx

Use the following command to verify if all the application resources got installed with their status:

kubectl get all -n multimodal-sample-app

Step 4: Copy the udf package for helm deployment#

DL Streamer Pipeline Server

To copy your own or existing model into DL Streamer Pipeline Server in order to run this sample application in Kubernetes environment:

The model package is available in the repository at edge-ai-suites/manufacturing-ai-suite/industrial-edge-insights-multimodal/configs/dlstreamer-pipeline-server/.

Copy the resources such as video and model from local directory to the to the dlstreamer-pipeline-server pod to make them available for application while launching pipelines.

cd edge-ai-suites/manufacturing-ai-suite/industrial-edge-insights-multimodal/configs/dlstreamer-pipeline-server/

POD_NAME=$(kubectl get pods -n multimodal-sample-app -o jsonpath='{.items[*].metadata.name}' | tr ' ' '\n' | grep deployment-dlstreamer-pipeline-server | head -n 1)

kubectl cp models $POD_NAME:/home/pipeline-server/resources/ -c dlstreamer-pipeline-server -n multimodal-sample-app

Time Series Analytics Microservice

To copy your own or existing model into Time Series Analytics Microservice in order to run this sample application in Kubernetes environment:

The following udf package is placed in the repository under edge-ai-suites/manufacturing-ai-suite/industrial-edge-insights-multimodal/configs/time-series-analytics-microservice.

- time-series-analytics-microservice/
    - models/
        - weld_anomaly_detector.cb
    - tick_scripts/
        - weld_anomaly_detector.tick
    - udfs/
        - requirements.txt
        - weld_anomaly_detector.py

Copy your new UDF package to the time-series-analytics-microservice pod:

cd edge-ai-suites/manufacturing-ai-suite/industrial-edge-insights-multimodal/configs/time-series-analytics-microservice # path relative to git clone folder
mkdir -p weld_anomaly_detector
cp -r models tick_scripts udfs weld_anomaly_detector/.

POD_NAME=$(kubectl get pods -n multimodal-sample-app -o jsonpath='{.items[*].metadata.name}' | tr ' ' '\n' | grep deployment-time-series-analytics-microservice | head -n 1)

kubectl cp weld_anomaly_detector $POD_NAME:/tmp/ -n multimodal-sample-app

Note: Run the commands only after performing the Helm install.

Step 5: Activate the Pipeline/UDF Deployment Package#

DL Streamer Pipeline Server

You use a Client URL (cURL) command to start the pipeline. Start this pipeline with the following cURL command.

curl -k https://localhost:30001/dsps-api/pipelines/user_defined_pipelines/weld_defect_classification -X POST -H 'Content-Type: application/json' -d '{
    "destination": {
        "metadata": {
            "type": "mqtt",
            "topic": "vision_weld_defect_classification"
        },
        "frame": {
            "type": "webrtc",
            "peer-id": "samplestream"
        }
    },
    "parameters": {
        "classification-properties": {
            "model": "/home/pipeline-server/resources/models/weld-defect-classification-f16-DeiT/deployment/Classification/model/model.xml",
            "device": "CPU"
        }
    }
}'

Time Series Analytics Microservice

NOTE: UDF inferencing on GPU is not supported.

Run the following command to activate the UDF deployment package:

curl -k -X 'GET' \
'https://localhost:30001/ts-api/config?restart=true' \
-H 'accept: application/json'

Step 6: Verify the Results#

Follow the verification steps in the Get Started guide

Uninstall Helm Charts#

To uninstall Helm charts:

helm uninstall multimodal-weld-defect-detection -n multimodal-sample-app
kubectl get all -n multimodal-sample-app # It may take a few minutes for all application resources to be cleaned up.

Configure Alerts in Time Series Analytics Microservice#

To configure alerts in Time Series Analytics Microservice, follow the steps.

Troubleshooting#

Check pod details or container logs to diagnose failures:

kubectl get pods -n multimodal-sample-app
kubectl describe pod <pod_name> -n multimodal-sample-app # Shows details of the pod
kubectl logs -f <pod_name> -n multimodal-sample-app # Shows logs of the container in the pod