kubernetes-the-hard-way/docs/01-container-runtime.md

7.8 KiB

Container runtime

In this section, we will focus on the container runtime.

image

runc

First of all, since Kubernetes is an orchestrator for containers, we would like to figure out how to run containers. A thing like OCI can help us here.

The OCI is a project under the Linux Foundation is aims to develop open industry standards for container formats and runtimes. The primary goal of OCI is to ensure container portability and interoperability across different platforms and container runtime implementations. The OCI has two main specifications, Runtime Specification (runtime-spec) and Image Specification (image-spec).

As we can see from the description - OCI is a standard that tells us what is a container image and how to run it.

But it is only a standard, obviously, there is some tool that implements this standard. And it is true, runc is a reference implementation of the OCI runtime specification.

So let's install it and run some container with the usage of runc

First of all we need to download runc binaries

wget -q --show-progress --https-only --timestamping \
  https://github.com/opencontainers/runc/releases/download/v1.2.6/runc.amd64

After the download process is complete, we need to move runc binaries to proper folder

mv runc.amd64 runc \
  && chmod +x runc \
  && mv runc /usr/local/bin/

Now, as we have runc configured, we can run busybox container

mkdir -p busybox-container/rootfs/bin \
  && cd busybox-container/rootfs/bin \
  && wget https://www.busybox.net/downloads/binaries/1.31.0-defconfig-multiarch-musl/busybox-x86_64 \
  && chmod +x busybox-x86_64 \
  && ./busybox-x86_64 --install . \
  && cd ./../.. \
  && runc spec \
  && sed -i 's/"sh"/"echo","Hello from container runned by runc!"/' config.json

In this step, we downloaded the busybox image, unarchived it, and created the proper files, required by runc to run the container (including container configuration and files that will be accessible from the container). So, let's run our container

runc run busybox

Output:

Hello from container runned by runc!

Great, we created our first container in this tutorial. Now we will clean up our workspace.

cd .. \
  && rm -r busybox-container

containerd

As we can see, runc can run containers, but runc interface is something unknown for kubernetes.

There is another standard defined which is used by kubelet to communicate with container runtime - CRI

The CRI is a plugin interface which enables the kubelet to use a wide variety of container runtimes, without having a need to recompile the cluster components.

In this tutorial, we will use containerd as a tool which is compattible with CRI.

To deploy containerd, first of all, we need to download it.

wget https://github.com/containerd/containerd/releases/download/v2.0.4/containerd-2.0.4-linux-amd64.tar.gz

After download process complete, we need to unzip and move containerd binaries to proper folder

mkdir containerd \
  && tar -xvf containerd-2.0.4-linux-amd64.tar.gz -C containerd \
  && mv containerd/bin/* /bin/

In comparison to the runc, containerd is a service that works like a service that can be called by someone to run a container. It means that we need to run it before we can start communicating with it.

We will configure containerd as a service.

To do that, we need to create containerd configuration file

{
mkdir -p /etc/containerd/

cat << EOF | tee /etc/containerd/config.toml
[debug]
  level = "debug"
  
[plugins]
  [plugins.'io.containerd.cri.v1.images']
    snapshotter = "native"
  [plugins."io.containerd.cri.v1.runtime"]
    [plugins."io.containerd.cri.v1.runtime".containerd]
      default_runtime_name = "runc"
      [plugins."io.containerd.cri.v1.runtime".containerd.runtimes]
        [plugins."io.containerd.cri.v1.runtime".containerd.runtimes.runc]
          runtime_type = "io.containerd.runc.v2"
          snapshotter = "native"

          [plugins."io.containerd.cri.v1.runtime".containerd.runtimes.runc.options]
            BinaryName = "/usr/local/bin/runc"
EOF
}

As we can see, we configured containerd to use runc (we installed before) to run containers.

After configuration file create, we need to create containerd service

cat <<EOF | tee /etc/systemd/system/containerd.service
[Unit]
Description=containerd container runtime
Documentation=https://containerd.io
After=network.target

[Service]
ExecStart=/bin/containerd
Restart=always
RestartSec=5
Delegate=yes
KillMode=process
OOMScoreAdjust=-999
LimitNOFILE=1048576
LimitNPROC=infinity
LimitCORE=infinity

[Install]
WantedBy=multi-user.target
EOF

And now, run it

systemctl daemon-reload \
  && systemctl enable containerd \
  && systemctl start containerd

To ensure that our service successfully started, run

systemctl status containerd

Output:

● containerd.service - containerd container runtime
     Loaded: loaded (/etc/systemd/system/containerd.service; enabled; vendor preset: enabled)
     Active: active (running) since Sat 2023-04-15 21:04:43 UTC; 59s ago
       Docs: https://containerd.io
    Process: 1018 ExecStartPre=/sbin/modprobe overlay (code=exited, status=0/SUCCESS)
   Main PID: 1031 (containerd)
      Tasks: 9 (limit: 2275)
     Memory: 22.0M
     CGroup: /system.slice/containerd.service
             └─1031 /bin/containerd
...

Now, we have containerd service running. It means that we can try to create some container.

To do that, we need the tool called ctr, which is distributed as part of containerd (which means that we already installed it during the installation of containerd).

First of all, we will pull busybox image

ctr images pull docker.io/library/busybox:latest

After the pull process is complete - check our image

ctr images ls

Output:

REF                              TYPE                                                      DIGEST                                                                  SIZE    PLATFORMS                                                                                                                          LABELS
docker.io/library/busybox:latest application/vnd.docker.distribution.manifest.list.v2+json sha256:b5d6fe0712636ceb7430189de28819e195e8966372edfc2d9409d79402a0dc16 2.5 MiB linux/386,linux/amd64,linux/arm/v5,linux/arm/v6,linux/arm/v7,linux/arm64/v8,linux/mips64le,linux/ppc64le,linux/riscv64,linux/s390x -

Now, let's start our container

ctr run --detach --snapshotter native docker.io/library/busybox:latest busybox-container sh -c 'while sleep 1; do echo "Hi"; done'

Output:

Hello from container runned by containerd!

As we can see we successfully started the container, now we can check its status

ctr containers ls

Output:

CONTAINER             IMAGE                               RUNTIME
busybox-container     docker.io/library/busybox:latest    io.containerd.runc.v2

But there is no info about the status, we can see it by reviewing tasks

ctr task ls

Output:

TASK                  PID        STATUS
busybox-container     2862580    RUNNING

As we can see our container is in the stopped state (because the command was successfully executed and the container stopped).

Now, let's clean up our workspace and go to the next section.

Stop running command

kill -9 $(ctr task ls | grep busybox | awk '{print $2}')

And remove the created container

ctr containers rm busybox-container

We can check that list of containers and tasks should be empty

ctr task ls \
 && ctr containers ls

We should receive an empty output

Next: Kubelet