Topics

#howto #k8s #k8s-cluster #kubernetes #kubernetes-cluster #kubernetes-environment #lxc-containers #lxd

Language breakdown

No language data available.

▸ Files click to expand

README

kubernetes-lxd

A step-by-step guide to get kubernetes running inside an LXC container.

This guide is an alternative to minikube which also offers a local kubernetes environment. The advantage of the LXC approach is that everything runs natively on the host kernel without any virtualization costs from a Virtual Machine. For example, minikube causes such high CPU usage on the host (see minikube issue #3207), that development is impaired. The downside is more setup work to get the kubernetes environment running, its administration costs, and lower isolation of the kubernetes cluster.

Below, you find a step-by-step guide to setup an LXC container and install kubernetes on it.

Lxc installation

Lxc is similar to docker but aims more to be an OS container instead of just application containers. To use it, install lxd and initialize it using lxd init. When prompted, answer the following questions:

- Would you like to use LXD clustering? (yes/no) [default=no]: - Do you want to configure a new storage pool? (yes/no) [default=yes]: yes -> dir (any directory based provider should work) - Would you like to connect to a MAAS server? (yes/no) [default=no]: - Would you like to create a new local network bridge? (yes/no) [default=yes]: - What should the new bridge be called? [default=lxdbr0]: - What IPv4 address should be used? (CIDR subnet notation, “auto” or “none”) [default=auto]: - What IPv6 address should be used? (CIDR subnet notation, “auto” or “none”) [default=auto]: - Would you like LXD to be available over the network? (yes/no) [default=no]: no - Would you like stale cached images to be updated automatically? (yes/no) [default=yes] - Would you like a YAML "lxd init" preseed to be printed? (yes/no) [default=no]:

Starting your LXC container

Before you can fire up your lxc container, you have to make sure to create /etc/subuid and /etc/subgid with the following entries:

root:100000:1000000000 :100000:1000000000

Run systemctl restart lxd to have LXD detect your new maps.

As the base system for our kubernetes we will use Debian and call the lxc machine k8s-lxc. Now create your kubernetes host machine with

lxc launch images:debian/stretch k8s-lxc

Run lxc list to ensure that your machine k8s-lxc is up and running.

Note: To get an overview over the supported base images, run

lxc image list images:

Usual lxc containers are quite restricted in their capabilities.

Because we need to run docker and kubernetes in the lxc container, it is required to give the container the capabilities to manage networking configuration and create cgroups. For that, run lxc config edit k8s-lxc and merge in the following settings:

config:
	  linux.kernelmodules: iptables,ip6tables,netlinkdiag,nf_nat,overlay
	  raw.lxc: "lxc.apparmor.profile=unconfined\nlxc.cap.drop= \nlxc.cgroup.devices.allow=a\nlxc.mount.auto=proc:rw
	    sys:rw"
	  security.privileged: "true"
	  security.nesting: "true"

- kernel_modules: depending on the kernel of your host system, you need to add further kernel modules here. The ones listed above are for networking and for dockers overlay filesystem. - raw.lxc: this allows the lxc container to configure certain system resources. - security.privileged and security.nesting: for a privileged container which may create nested cgroups

Restart your lxc container. Unfortunately, lxc stop k8s-lxc does not work for me. I need to do lxc exec k8s-lxc reboot.

Using docker and kubernetes on zfs backed host systems

If your host system is backed by ZFS storage (e.g. an option for Proxmox), some adaption need to be made. ZFS currently lacks full namespace support an thus a dataset cannot be reached into a LXC container retaining full control over the child datasets. The easiest solution is to create two volumes for /var/lib/docker and /var/lib/kubelet and format these ext4.

zfs create -V 50G mypool/my-dockervol zfs create -V 5G mypool/my-kubeletvol mkfs.ext4 /dev/zvol/mypool/my-dockervol mkfs.ext4 /dev/zvol/mypool/my-kubeletvol One then just needs to reach in the two volumes at the right location. The configuration for Proxmox looks like this:

... mp0: /dev/zvol/mypool/my-dockervol,mp=/var/lib/docker,backup=0 mp1: /dev/zvol/mypool/my-kubeletvol,mp=/var/lib/kubelet,backup=0 ...

Installing kubernetes in the lxc container

Below, some commands will need to be executed inside the lxc container and others on the host.

$-prefix means to be executed on the host machine
@-prefix means to be executed inside the lxc container
no prefix means it does not matter where the command is executed

First ensure on your host system that $ cat /proc/sys/net/bridge/bridge-nf-call-iptables returns 1.

This is required by kubernetes but cannot be validated inside the lxc container.

On your host, ensure that $ conntrack -L produces some output (this means it works). If this requires additional kernel modules to be loaded, add those to the lxc container config. For example you might need to add in $ lxc config edit k8s-lxc

config:      linux.kernelmodules: xtconntrack,...

After that, verify that @ conntrack also works inside your lxc container. To enter your container as root, do $ lxc exec k8s-lxc /bin/bash.

Recent kubernetes versions want to read from /dev/kmsg which is not present in the container.

You need to instruct systemd to always create a symlink to /dev/console instead:

@ echo 'L /dev/kmsg - - - - /dev/console' > /etc/tmpfiles.d/kmsg.conf

This solution can cause infinite CPU usage in some cases; some(?) versions of systemd-journald read from /dev/kmsg and write to /dev/console, and if they're symlinked together, this will cause an infinite loop. If this affects you, link /dev/null to /dev/kmsg instead:

@ echo 'L /dev/kmsg - - - - /dev/null' > /etc/tmpfiles.d/kmsg.conf

Install docker and kubernetes runtime in the lxc container.

The following commands add the required repositories, install kubernetes with dependencies, and pin the kubernetes & docker version:

@ apt-get install apt-transport-https ca-certificates curl gnupg2 software-properties-common
   @ curl -fsSL https://download.docker.com/linux/debian/gpg | apt-key add -
   @ curl -s https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add -
   @ add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/debian $(lsb_release -cs) stable"
   @ add-apt-repository "deb http://apt.kubernetes.io/ kubernetes-xenial main"
   @ apt-get update
   @ apt-get install -y docker-ce kubelet kubeadm kubectl
   @ apt-mark hold kubelet kubeadm kubectl docker-ce

Configure the kubelet in the lxc container:

@ kubeadm init --ignore-preflight-errors=FileContent--proc-sys-net-bridge-bridge-nf-call-iptables
   @ kubeadm init phase addon all

For the first command you need to ignore the bridge-nf-call-iptables check which you have done manually before. In case you obtain an error like failed to parse kernel config in the preflight check, copy your host kernel config to from /boot to your lxc-guest /boot.

Disable the software container network infrastructure, because it is not needed for a dev environment:

@ sed -i 's/--network-plugin=cni //' /var/lib/kubelet/kubeadm-flags.env
   @ systemctl daemon-reload
   @ systemctl restart kubelet

(Optional) Reduce the replica count of the coreDNS service to 1.

@ kubectl scale -n kube-system deployment --replicas 1 coredns

Congratulations, if the last command worked, you now have kubernetes running in your lxc container.

Configure the host for working with k8s-lxc

On your host machine, add a host entry to access your k8s-lxc container by DNS name.

Find out the IP of your lxc container by running $ lxc list k8s-lxc. Add its IP in /etc/hosts

/etc/hosts
   <k8s-lxc-ip>   k8s-lxc

After that, it should be possible to ping the container with ping k8s-lxc.

Make your docker daemon in the lxc cluster available from your host.

There are two options. - insecure without authentication

Open docker in the lxc container so that it can be accessed from outside. Run

@ systemctl edit docker.service

and add the following lines:

[Service]      ExecStart=      ExecStart=/usr/bin/dockerd -H fd:// -H tcp://0.0.0.0:2376

This is necessary, because the default unit definition also defines a host address, so that it cannot be overridden by the configuration file.

Restart docker in the lxc container, so that the changes take effect. Run

@ systemctl restart docker.service

On your host machine, you can then talk to this docker daemon by setting the environment variable

$ export DOCKER_HOST="tcp://k8s-lxc:2376"      $ docker ps   # should show some kubernetes infrastructure containers

- secure with authentication

1. In your lxc container, change into folder /etc/docker. Follow the instructions from docker.com to set up server and client credentials. 2. Then create a /etc/docker/daemon.json with the following content:

@ cat <<-EOF > /etc/docker/daemon.json           {              "tls": true,              "tlscacert": "/etc/docker/ca.pem",              "tlscert": "/etc/docker/server-cert.pem",              "tlskey": "/etc/docker/server-key.pem",              "hosts": ["fd://", "tcp://0.0.0.0:2376"]           }           EOF

3. Edit your docker.service unit by running @ systemctl edit docker.service and add/change the following:

[Service]           ExecStart=           ExecStart=/usr/bin/dockerd

This is necessary, because the default unit definition also defines a host address, so that it cannot be overridden by the configuration file.

4. Pull the certificate and client keys to a config directory

$ mkdir ~/.docker-lxc && cd ~/.docker-lxc          $ lxc file pull k8s-lxc/etc/docker/ca.pem .          $ lxc file pull k8s-lxc/etc/docker/cert.pem .          $ lxc file pull k8s-lxc/etc/docker/key.pem .

If you do not plan to use docker on your host as well, you can also use the default docker config directory ~/.docker/.

5. Set the following environment variables to access the docker daemon in the lxc cluster

export DOCKERTLSVERIFY="1"          export DOCKERCERTPATH="~/.docker-lxc/"          export DOCKER_HOST="tcp://k8s-lxc:2376"          export DOCKERAPIVERSION="1.37"

Set up a kubectl context on your host system to talk to your kubernetes installation in the lxc container:

$ lxc file pull k8s-lxc/etc/kubernetes/admin.conf .    $ KUBECONFIG=~/.kube/config:admin.conf kubectl config view --raw > config.tmp    $ mv config.tmp ~/.kube/config    $ kubectl config use-context k8s-lxc-admin@k8s-lxc

The second line does some magic to merge the admin access credentials into the existing KUBECONFIG file.

Kubeadm usually taints the master node which prevents non-system pods from being scheduled.

If this applies to you, do

kubectl edit node

and remove the taint.

Kubernetes fine-tuning

If needed, configure an ingress controller for your kubernetes installation

- Ensure that IP forwarding is enabled on your host

sysctl net.ipv4.ip_forward

or enable it permanently with

sysctl -w net.ipv4.ip_forward=1

- Create the ingress controller

kubectl apply -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/master/deploy/static/mandatory.yaml

- To access the ingress via the default http/https ports, add hostPort directives to its deployment template. Run kubectl edit -n ingress-nginx deployment nginx-ingress-controller and change the port definitions to

ports:      - containerPort: 80        hostPort: 80        name: http        protocol: TCP      - containerPort: 443        hostPort: 443        name: https        protocol: TCP

Because we are skipping the ingress-nginx service, you should also remove --publish-service=$(POD_NAMESPACE)/ingress-nginx from the arguments to nginx-ingress-controller.

Disable leader election for control plane components, because this it is obsolete for a single node deployment.

sed -i 's/--leader-elect=true/--leader-elect=false/' \
      /etc/kubernetes/manifests/{kube-controller-manager.yaml,kube-scheduler.yaml}

(Optional) Create an SSL certificate for your lxc container to secure traffic

- Follow the instructions from kubernetes.io. Using the following commands:

# prepare certificate signing request cat <<EOF | cfssl genkey - | cfssljson -bare server { "hosts": [ "k8s-lxc" ], "CN": "k8s-lxc", "key": { "algo": "ecdsa", "size": 256 } } EOF # create certificate signing request cat <<EOF | kubectl create -f - apiVersion: certificates.k8s.io/v1beta1 kind: CertificateSigningRequest metadata: name: k8s-lxc-csr spec: groups: - system:authenticated request: $(cat server.csr | base64 | tr -d '\n') usages: - digital signature - key encipherment - server auth EOF # approve the certificate, using the k8s cluster ca.pem kubectl certificate approve k8s-lxc-csr # retrieve the server certificate kubectl get csr k8s-lxc-csr -o js \ | base64 --decode > server.crt

# create tls secret kubectl create secret tls k8s-lxc --cert=server.crt --key=server-key.pem

(Optional) Secure access to your cluster by creating a user with edit rights.

Use the script src/setup-default-user.sh to set up authentication by client certificate for a user with name default-user.

Check if everything is running correctly by deploying a small test application

kubectl apply -f src/test-k8s-lxc.yaml

You should be able to access the application from your browser http://k8s-lxc.

If that does not work, try to access the test service from within your kubernetes network.

$ kubectl run --generator=run-pod/v1 -ti --image nicolaka/netshoot curl    > curl test # should fetch a minimal greeting    Ctrl-D    $ kubectl delete pod curl