Skip to content

Installation.md

Latest commit

 

History

History
96 lines (79 loc) · 5.42 KB

Installation.md

File metadata and controls

96 lines (79 loc) · 5.42 KB
Raw

Installation Guides

Awareness Engine can be deployed in various ways. In this section all necessary steps for every variant are described. So far the following scenarios are supported:

  • Deployment on Kubernetes clusters (recommended productive mode)
  • Docker Compose
  • Installation via APT packages (embedded deployment)

Please note, that regardless of deployment method - SAE works in all of them with the same general concepts described here.

Installation on Kubernetes cluster

  1. Set up K3s cluster \
    • Create K3s config file at /etc/rancher/k3s/config.yaml (create directory if it does not exist) 
      write-kubeconfig-mode: "0644"
tls-san:
    - "TAILSCALE_HOSTNAME" # <-- Adapt this value
disable:
    - servicelb
    - traefik
      • tls-san only has to be set if the hostname the cluster is supposed to be managed through is not equal to the actual hostname (e.g. it is on Tailscale)
      • As we currently do not open any ports and do not host publicly host any services traefik and servicelb can be disabled
      • --write-kubeconfig-mode 644 enables managing the cluster through any user on the machine
    • Install K3s curl -sfL https://get.k3s.io | sh -s - (use the same command for upgrading, assuming you have the config file in place)
    • If the machine you install K3s on has a NVIDIA GPU available and wants to use it for running the pipeline, you have to follow the instructions in here
  2. Install Timescale DB (if do not have one on another machine and you want to store the pipeline output data)
    • Create a database for pipeline data and a table / hypertable. See here for instructions.
  3. Install SAE
    • Copy the values template file and adapt it to your needs (see the file for information) 
      cp ./helm/sae/customvalues.template my.values.yaml
    • Install helm chart 
      helm install <release-name> oci://registry-1.docker.io/starwitorg/sae -f my.values.yaml
  4. Set up and use the watch tool (see here) to inspect the pipeline visually
  5. Check out the internal pipeline metrics
    • Forward Grafana port to your local machine and log in (user admin, password admin)

Docker Compose

Docker Compose is described in detail in this section.

APT/Debian package installation

On some embedded devices Docker may not be an option, so SAE components are also shipped into installable Debian packages. So far embedded installation targets at running all necessary SAE components, such that on a device tracked objects are geo-mapped.

This way application developers can decide, whether to use collected data, to implement function directly in hardware, or to forward aggregated information to a receiving backend. See section sample use cases for more details.

Packages are published as release in every component. This table contains links to each package.

Component Repository / URI
Video Source https://github.com/starwit/video-source-py/releases
Object Detector https://github.com/starwit/object-detector/releases
Object Tracker https://github.com/starwit/object-tracker/releases
Geo-Mapper https://github.com/starwit/geo-mapper/releases
Redis Writer https://github.com/starwit/sae-redis-writer/releases

For future release setting up a proper package source is under consideration.

Installation steps

The following installation steps are the same for all Linux derivates. For distro-specific steps see section(s) below.

  • Install ValKey (see here for help)
  • Download *.deb file for each component
  • Create folder /etc/starwit and place config files (see next section)
  • Install all component (see above table for order) using: 
    apt install ./package_for_component.deb

You can skip config file step, but then each component will install a config file with default values - which needs to be adapted.

Configuration

The following table is an overview of all configuration files

Component Example Config File Notes
Video Source settings.yaml video sources coordinates (RTSP, usename/password)
Object Detector settings.yaml Select detection model, inferencing size,...
Object Tracker settings.yaml Select tracking method
Geo-Mapper settings.yaml Provide geo coordinates of camera
Redis Writer settings.yaml Transfer data to backend

Nvidia Jetson

# Add Nvidia Sources
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/arm64/cuda-keyring_1.1-1_all.deb
sudo apt install ./cuda-keyring_1.1-1_all.deb

# avoid possible version conflict
sudo apt remove python3-sympy