Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 4 Next »

Introduction

This tutorial is going to explain how to setup a local Kubernetes cluster and minimal helm setup to run and deploy SDC (but can be extended to several/all ONAP components) on a single host.

The rationale for this tutorial is to spin up a development environment quickly and efficiently without the hassle of setting up a multi node cluster/Network file share that are needed in a full deployment.

This was a trial in G release but is now extended to the requirements of H release.


WARNING :

There are limitations in using this setup, the purpose is to allow for developers to quickly get access to a working environment 'the way their component is supposed to be deployed'

This is not meant for production obviously, and the tweaks that are done to the OOM/K8s setup are most likely going to evolve with further releases.

Minimum Requirements

  • One VM running Ubuntu 20.04 LTS (should also work on 18.04), with internet access to download charts/containers and the oom repo
  • Root/sudo privileges
  • Sufficient RAM depending on how many components you want to deploy
    • Around 20G of RAM allows for a few components, the minimal setup for SDC is enabling :
      • Shared Cassandra
      • AAF
      • Portal (if you need UI access)
      • SDC
    • This was tested with a huge VM - 128G of RAM and 12 VCPU, running most of the components in Honolulu development.
    • This was tested with a small VM to run components on a local Laptop (need enough ram to create a 20G Ram VM using VBox, VMWare...)
  • above 160G available storage should be sufficient, mostly depends on how many components you want to enable in OOM charts.
    • Storage is required mostly to store container images

Overall Procedure

  • Install/remove Microk8s with appropriate version
  • Install/remove Helm with appropriate version
  • Tweak Microk8s
  • Download oom repo
  • Install the needed Helm plugins
  • Install ChartMuseum as a local helm repo
  • Install docker (now needed to build oom charts)
  • Build all oom charts and store them in the chart repo
  • Tweak oom override file to fine tune deployment based on your VM capacity and component needs
  • Deploy charts
  • Enable UI access

1) Install/Upgrade Microk8s with appropriate version

Why Microk8s ?

Microk8s is a bundled lightweight version of kubernetes maintained by Canonical, it has the advantage to be well integrated with snap on Ubuntu, which makes it super easy to manage/upgrade/work with

More info on : https://microk8s.io/docs

There are 2 things to know with microk8s :

1) it is wrapped by snap, which is nice but you need to understand that it's not exactly the same as having a proper k8s installation (more info below on some specific commands)

2) it is not using docker as the container runtime, it's using containerd, it's not an issue, just be aware of that as you won't see containers using classic docker commands


How to remove/install/remove microk8s ?

If you have a previous version of microk8s, you first need to uninstall it (upgrade is possible but it is not recommended between major versions so I recommend to uninstall as it's fast and safe)

sudo snap remove microk8s

You need to select the appropriate version to install, to see all possible version do :

sudo snap info microk8s

this tutorial is focused on Honolulu release so we will use k8s version 1.19, to do so, you just need to select the appropriate channel

sudo snap install microk8s --classic --channel=1.19/stable

You may need to change your firewall configuration to allow pod to pod communication and pod to internet communication :

sudo ufw allow in on cni0 && sudo ufw allow out on cni0
sudo ufw default allow routed

Addons ?

MicroK8s is lightweight but comes with several possible addons, OOM and ONAP requires just a few to be enabled, you can choose to enable more if you want to

DNS addon : we need the DNS addon so that pods can 'see' each other by host name.

Storage addon : we will enable the default Host storage class, this allows local volume storage that are used by some pods to exchange folders between containers.

microk8s enable dns storage

that's it, you should have a running k8s cluster, ready to host ONAP pods

I recommend to get familiar with microk8s, here are a few useful commands but you can read more on the microk8s website :

  • microk8s status: Provides an overview of the MicroK8s state (running / not running) as well as the set of enabled addons
  • microk8s enable: Enables an addon
  • microk8s disable: Disables an addon
  • microk8s kubectl: Interact with kubernetes
  • microk8s config: Shows the kubernetes config file
  • microk8s inspect: Performs a quick inspection of the MicroK8s intallation
  • microk8s reset: Resets the infrastructure to a clean state → very useful for a dev lab
  • microk8s stop: Stops all kubernetes services
  • microk8s start: Starts MicroK8s after it is being stopped

2) Install/remove Helm with appropriate version

Helm is the package manager for k8s, we require a specific version for each ONAP release, the best is to look at the OOM guides to see which one is required (link to add)

For the Honolulu release we need Helm 3 - A significant improvement with Helm3 is that it does not require a specific pod running in the kubernetes cluster (no more Tiller pod)

As Helm is self contained, it's pretty straightforward to install/upgrade

I recommend putting helm in local bin folder as a softlink, this way it's easy to switch between versions if you need to

 wget https://get.helm.sh/helm-v3.5.2-linux-amd64.tar.gz
 tar -zxvf helm-v3.5.2-linux-amd64.tar.gz
 sudo mv linux-amd64/helm /usr/local/bin/helm-v3.5.2
 sudo ln -s /usr/local/bin/helm-v3.5.2 /usr/local/bin/helm

3) Tweak Microk8s

The below tweaks are not strictly necessary, but they help in making the setup more simple and flexible.

A) Increase the max number of pods

As ONAP may deploy a significant amount of pods, we need to inform kubelet to allow more than the basic configuration (as we plan an all in box setup), if you only plan to run a limited number of components, this is not need

to change the max number of pods, we need to add a parameter to the startup line of kubelet

edit the file located at

/var/snap/microk8s/current/args/kubelet

add the following line at the end :

--max-pods=250

save the file and restart kubelet to apply the change :

sudo service snap.microk8s.daemon-kubelet restart


B) run a local copy of kubectl

Microk8s comes bundled with kubectl, you can interact with it by doing:

microk8s kubectl describe node

to make things simpler as we will most likely interact a lot with kubectl, let's install a local copy of kubectl so we can use it to interact with the kubernetes cluster in a more straightforward way

We need kubectl 1.19 to match the cluster we have installed

curl -LO https://storage.googleapis.com/kubernetes-release/release/v1.19.7/bin/linux/amd64/kubectl
chmod +x ./kubectl
sudo mv ./kubectl /usr/local/bin/kubectl

If you want to use a the same link trick as for helm (see above) this could allow you to switch between kubectl versions if needed

Now we need to provide our local kubectl client with a proper config file so that it can access the cluster, microk8s allows to retrieve the cluster config very easily

Simply create a .kube folder in your home directory and dump the config there

cd
mkdir .kubecd .kube
microk8s.config > config
chmod 700 config

the last line is there to avoid helm complaining about too open permission

you should now have helm and kubectl ready to interact with each other, you can verify this by trying :

kubectl version

this should output both the local client and server version

Client Version: version.Info{Major:"1", Minor:"19", GitVersion:"v1.19.7", GitCommit:"1dd5338295409edcfff11505e7bb246f0d325d15", GitTreeState:"clean", BuildDate:"2021-01-13T13:23:52Z", GoVersion:"go1.15.5", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"19+", GitVersion:"v1.19.7-34+02d22c9f4fb254", GitCommit:"02d22c9f4fb2545422b2b28e2152b1788fc27c2f", GitTreeState:"clean", BuildDate:"2021-02-11T20:13:16Z", GoVersion:"go1.15.8", Compiler:"gc", Platform:"linux/amd64"}

4) Download OOM repo

OOM provides a couple of plugins that we need to install for helm, the easiest is to get them straight from the repository.

Since we will also build the charts locally, it's the right time to download a local copy, you'll need to install git if it's not available yet on your machine

the example below pulls the latest version from master, it's probably wiser to select the right version (honolulu branch or a specific review you want to test)

cd
git clone --recursive "https://gerrit.onap.org/r/oom"

5) Install the needed Helm plugins

Onap deployments are using the deploy and undeploy plugins for helm

to install them just run :

helm plugin install ./oom/kubernetes/helm/plugins/undeploy/
helm plugin install ./oom/kubernetes/helm/plugins/deploy/

this will copy the plugins into your home directory .helm folder and make them available as helm commands

Another plugin we need is the push plugin, with helm3 there is no more an embedded repo to use.

helm plugin install https://github.com/chartmuseum/helm-push.git

Once all plugins are installed, you should see them as available helm commands when doing :

helm --help

6) Install the charmuseum repository

To align with how the previous release were deployed, we will setup a local chart repository.

To do, download the chartmuseum script and run it as a background task

curl https://raw.githubusercontent.com/helm/chartmuseum/main/scripts/get-chartmuseum | bash
mkdir ~/chartmuseum
chartmuseum --port=8080   --storage="local"   --storage-local-rootdir="~/chartstorage" &

you should see the chartmuseum repo starting locally, you can press enter to come back to your terminal

you can now inform helm that a local repo is available for use :

helm repo add local http://localhost:8080


Note this is a very basic setup, meaning you'll need to startup your local repo if you shut down your VM or if you exit this terminal, there are other ways to do it, like using a docker compose to ensure 'it's always on'


7) Install docker

During the processing of the oom helm charts (oom-cert-service component of platform), docker is needed as it is used to create certificates.

Follow the docker install instructions for Ubuntu here : https://docs.docker.com/engine/install/ubuntu/

unless you already have docker, in which case you can skip this part altogether.


8) Build all oom charts and store them in the chart repo

You should be ready to build all helm charts, go into the oom/kubernetes folder and run a full make

cd ~/oom/kubernetes
make all

You can speed up the make skipping the linting of the charts

make all -e SKIP_LINT=TRUE

You'll notice quite a few message popping into your terminal running the chartmuseum, showing that it accepts and store the generated charts, that's normal, if you want, just open another terminal to run the helm commands

Once the build completes, you should be ready to deploy ONAP


9) Tweak oom override file to fine tune deployment based on your VM capacity and component needs

Before deploying, it's good to decide which ONAP components you need and create an override file. The override file will tune your deployment to your specific needs.

We have a few considerations to take on our stand alone VM :

1) CPU/Memory might be limited, so be careful on what you enable / disable

2) Timeouts : the more component you enable, the more the liveness probes and readiness probes timeouts may have an influence on how often K8S





  • No labels