This guide outlines the steps to deploy a development environment comprised of hardware-observer and COS on a single physical machine.
Most of hardware observer's functionality can only be experienced when installing the charm on server-grade hardware and relating to a COS installation. As this can be a relatively expensive setup, we are offering this guide to document how to create an environment for development or rapid prototyping on a single physical host. Note that the proposed topology is in no way appropriate for production, nor it can be considered a supported deployment.
If you have a physical machine that you can fully take over and expect to redeploy when you're done, you can follow these steps to set up the environment.
A physical machine. Here, "physical machine" refers to machines that are not VMs or containers and have access to real hardware resources like RAID cards and BMC management tools.
A juju controller. You can find what is juju and how to deploy it here
First, you need to bootstrap a Juju controller on a machine. For simplicity and convenience in this guide, we will use a default LXD controller. It's important to note that you can use any machine within the same network as the one you plan to deploy the hardware-observer on to serve as a controller. However, since we are utilizing only one physical machine in this setup, we will bootstrap an LXD controller on it.
Due to some historical reasons, the machine hosting the LXD controller cannot be added to the model unless the default LXD bridge is renamed.
lxc network create br0 -t bridge
lxc profile edit default # Rename lxdbr0 to br0 in the profile
lxc network delete lxdbr0
Bootstrap a LXD controller on juju:
juju bootstrap localhost lxd-controller
To add the physical machine to the model, which is managed by a Juju controller on an LXD cloud, you need to allow the juju client to log into it via SSH as a user with sudo rights. On an Ubuntu machine, the ubuntu user typically satisfies this requirement:
cat ~/.local/share/juju/ssh/juju_id_rsa.pub >> ~/.ssh/authorized_keys
Now you can create a model and add your physical machine via the manual provider:
juju add-model hw-obs
# Use a different username if needed
# It is recommended to use the IP address of br0 for more reliable operation
BR0_ADDR=$(ip -4 -j a sho dev br0 | jq -r .[].addr_info[0].local)
juju add-machine ssh:ubuntu@$BR0_ADDR
juju switch hw-obs
juju deploy ubuntu --to 0
juju deploy hardware-observer
juju deploy grafana-agent
Add necessary relations:
juju relate hardware-observer ubuntu
juju relate grafana-agent ubuntu
juju relate grafana-agent hardware-observer
Set up microk8s. Steps can be found in this guide.
Install microK8s package, it is required to use a strictly confined version:
sudo snap install microk8s --channel 1.30-strict
Add your user to the microk8s group for unprivileged access:
sudo adduser $USER snap_microk8s
Give your user permissions to read the ~/.kube directory:
sudo chown -f -R $USER ~/.kube
Wait for microK8s to finish initialising:
sudo microk8s status --wait-ready
Enable the 'storage' and 'dns' addons:
sudo microk8s enable hostpath-storage dns
The COS bundle comes with Traefik to provide ingress, for which the metallb addon should be enabled:
# The IP address 2.2.2.2 is arbitrary and is used to determine the preferred source IP address for routing.
IPADDR=$(ip -4 -j route get 2.2.2.2 | jq -r '.[] | .prefsrc')
sudo microk8s enable metallb:$IPADDR-$IPADDR
Alias kubectl so it interacts with microK8s by default:
sudo snap alias microk8s.kubectl kubectl
Ensure your new group membership is apparent in the current terminal: (Not required once you have logged out and back in again)
newgrp snap_microk8s
Juju recognises a local microK8s cloud automatically, bootstrap a microk8s controller:
juju bootstrap microk8s k8s-controller
Before deploying the COS bundle, wait for all the microk8s addons to be rolled out:
microk8s kubectl rollout status deployments/hostpath-provisioner -n kube-system -w
microk8s kubectl rollout status deployments/coredns -n kube-system -w
microk8s kubectl rollout status daemonset.apps/speaker -n metallb-system -w
Create a COS model and deploy COS-lite bundle:
juju add-model cos
juju switch cos
curl -L https://raw.githubusercontent.com/canonical/cos-lite-bundle/main/overlays/offers-overlay.yaml -O
juju deploy cos-lite --trust --overlay ./offers-overlay.yaml
Switch back to your physical model and add the relations to COS:
juju switch hw-obs
juju relate grafana-agent k8s-controller:cos.prometheus-receive-remote-write
juju relate grafana-agent k8s-controller:cos.grafana-dashboards
juju relate grafana-agent k8s-controller:cos.loki-logging
Go to COS model and run:
juju run grafana/0 get-admin-password
This command will show the grafana dashboard endpoint and default password. The default username is "admin". Now you should be able to access grafana.