Docker Installation: Difference between revisions

From WilliamsNet Wiki
Jump to navigation Jump to search
mNo edit summary
mNo edit summary
Tag: visualeditor
 
(8 intermediate revisions by 2 users not shown)
Line 1: Line 1:
__NOTOC__
= Docker Community Edition install =
This installs the official Docker distrubution from the '''community edition''' repository.
This installs the official Docker distrubution from the '''community edition''' repository.


__NOTOC__
== Fedora ==
== Fedora ==
The docker team officially supports Fedora, but the rapid pace of Fedora releases (every 6 months) and the changing cgroups support (fedora has moved to v2 but docker is still on v1) has led to delays with the Fedora 33 release of docker-ce.  I expect this will happen again for future releases, so the instructions below accommodate this possibility.
The docker team officially supports Fedora, but the rapid pace of Fedora releases (every 6 months) and the changing cgroups support (fedora has moved to v2 but docker is still on v1) has led to delays with the Fedora 33 release of docker-ce.  I expect this will happen again for future releases, so the instructions below accommodate this possibility.
Line 23: Line 24:


=== Prerequisites ===
=== Prerequisites ===
* [[Basic CentOS 7 Installation]] or [[Basic CentOS 8 Installation]]
* [[Basic CentOS 7 Installation]]
* [[CUDA + NVIDIA drivers]] (if GPU is present)
* [[CUDA + NVIDIA drivers]] (if GPU is present)


Line 29: Line 30:
  sudo yum install -y yum-utils device-mapper-persistent-data lvm2
  sudo yum install -y yum-utils device-mapper-persistent-data lvm2
  sudo yum-config-manager --add-repo <nowiki>https://download.docker.com/linux/centos/docker-ce.repo</nowiki>
  sudo yum-config-manager --add-repo <nowiki>https://download.docker.com/linux/centos/docker-ce.repo</nowiki>
=== Install containerd.io (CentOS 8 only) ===
RHEL/CentOS 8 replaced the normal docker with podman/buildah, so we have to manually install the containerd.io runtime.  Choose the latest version of the RPM from the CentOS 7 repo. (see https://linuxconfig.org/how-to-install-docker-in-rhel-8)
dnf install https://download.docker.com/linux/centos/7/x86_64/stable/Packages/containerd.io-1.2.10-3.2.el7.x86_64.rpm


=== Install Docker ===
=== Install Docker ===
Line 41: Line 38:


=== Prerequisites ===
=== Prerequisites ===
* [[Basic Debian 10 Installation]]
* [[Basic Debian Installation]]


=== Set up Docker-CE Repository ===
=== Set up Docker-CE Repository ===
Line 51: Line 48:
=== Install Docker ===
=== Install Docker ===
  sudo apt-get install -y docker-ce docker-ce-cli containerd.io
  sudo apt-get install -y docker-ce docker-ce-cli containerd.io
== Configuring Container Runtime ==
Kubernetes says that it prefers that docker use 'systemd' as the cgroupdriver ([https://kubernetes.io/docs/setup/production-environment/container-runtimes/#cgroup-drivers see k8s docs here]).  To accomplish this, add this to the /etc/docker/daemon.json file:
{
  "exec-opts": ["native.cgroupdriver=systemd"],
  "log-driver": "json-file",
  "log-opts": {
    "max-size": "100m"
  },
  "storage-driver": "overlay2",
  "storage-opts": [
    "overlay2.override_kernel_check=true"
  ]
}
Note that this must be integrated with any changes needed to implement the GPU runtime (see below).


== Finishing the Install ==
== Finishing the Install ==
Line 88: Line 68:
  scp -r aslan:.docker ~  
  scp -r aslan:.docker ~  


=== GPU Nodes ===
= GPU Nodes =
On GPU-enabled nodes, install the nvidia runtime.
On GPU-enabled nodes, install the [[NVIDIA Docker Runtime]].
 
(originally from https://github.com/NVIDIA/nvidia-docker)
<pre>sudo yum-config-manager --add-repo https://nvidia.github.io/nvidia-docker/centos7/x86_64/nvidia-docker.repo
sudo yum install -y nvidia-docker2
sudo systemctl restart docker </pre>
 
Test nvidia-smi with the latest official CUDA image
docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi
 
Make the nvidia runtime default
* add entry to /etc/docker/daemon.json
* note that this gets reset when docker updates
* resulting daemon.json file looks like this:
 
<pre>{
    "runtimes": {
        "nvidia": {
            "path": "/usr/bin/nvidia-container-runtime",
            "runtimeArgs": []
        }
    },
    "default-runtime": "nvidia"
} </pre>
 
... or copy from aslan
scp aslan:/etc/docker/daemon.json /etc/docker
 
Reload the configuration to enable the change
sudo systemctl restart docker
 
Test the GPU performance using a simple NVIDIA GPU Cloud container with the CUDA nbody sample program
docker run -it --rm nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -benchmark -fp64
 
==== Reloading Repository Certificates ====
 
Rather frequently, it seems, the NVIDIA folks invalidate their signing certificates for their repositories. When that happens, you just need to delete the certificates and let the 'yum' command reload them on the next update:
 
$(sed -n 's/releasever=//p' /etc/yum.conf)
${DIST:-$(. /etc/os-release; echo $VERSION_ID)}
sudo rpm -e gpg-pubkey-f796ecb0
sudo gpg --homedir /var/lib/yum/repos/$(uname -m)/$DIST/nvidia-container-runtime/gpgdir --delete-key f796ecb0
sudo gpg --homedir /var/lib/yum/repos/$(uname -m)/$DIST/libnvidia-container/gpgdir --delete-key f796ecb0
sudo gpg --homedir /var/lib/yum/repos/$(uname -m)/$DIST/nvidia-docker/gpgdir --delete-key f796ecb0
sudo yum -y makecache

Latest revision as of 12:24, 17 October 2021

Docker Community Edition install[edit]

This installs the official Docker distrubution from the community edition repository.

Fedora[edit]

The docker team officially supports Fedora, but the rapid pace of Fedora releases (every 6 months) and the changing cgroups support (fedora has moved to v2 but docker is still on v1) has led to delays with the Fedora 33 release of docker-ce. I expect this will happen again for future releases, so the instructions below accommodate this possibility.

Prerequisites[edit]

Set up Docker-CE Repository[edit]

sudo dnf config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
dnf list docker\*

If you don't see the docker-ce* packages in the listing, you need to enable the test repository (CAUTION): 

sudo dnf config-manager --set-enabled docker-ce-test

Then you install the packages 

sudo dnf install docker-ce docker-ce-cli containerd.io

CentOS[edit]

( originally from https://docs.docker.com/engine/installation/linux/centos/ )

Prerequisites[edit]

Set up Docker-CE Repository[edit]

sudo yum install -y yum-utils device-mapper-persistent-data lvm2
sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo

Install Docker[edit]

sudo yum install -y docker-ce docker-ce-cli containerd.io

Debian[edit]

( originally from https://docs.docker.com/engine/installation/linux/debian/ )

Prerequisites[edit]

Set up Docker-CE Repository[edit]

sudo apt-get install -y apt-transport-https ca-certificates curl gnupg2 software-properties-common
curl -fsSL https://download.docker.com/linux/debian/gpg | sudo apt-key add -
sudo add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/debian $(lsb_release -cs) stable"
sudo apt-get update

Install Docker[edit]

sudo apt-get install -y docker-ce docker-ce-cli containerd.io

Finishing the Install[edit]

Make sure the docker daemon is running and set up to start automatically (should already be done by the package install): 

sudo systemctl enable --now docker

Test installation 

sudo docker run --rm hello-world
sudo docker image rm -f hello-world

Enable standard user access to docker commands (requires $USER to log out and back in to activate) 

sudo groupadd docker
sudo usermod -aG docker $USER

In order to access the local repositories, we need to copy the certs into the local docker config 

sudo scp -r aslan:/etc/docker/certs.d /etc/docker

Copy over the .docker directory from aslan for both root and user(s) to get login credentials 

sudo scp -r aslan:.docker /root
scp -r aslan:.docker ~

GPU Nodes[edit]

On GPU-enabled nodes, install the NVIDIA Docker Runtime.

Retrieved from "http://wiki.williams-net.org/index.php?title=Docker_Installation&oldid=560"