Linux dependencies

Note: Don't follow this step if you plan to run BlurIt OP on Windows. Go there instead.

Choice 1 : use the Blurit CLI (recommended)

The install_dependencies.sh script shipped with older archives is deprecated. Dependencies are now installed by the Blurit CLI, with the blurit requirements command described below.

Step 1 : extract the archive

The Blurit CLI enables you to easily configure, start, and interact with Blurit. It provides a command-line interface to manage features and streamline platform usage.

The blurit binary is provided in the archive you received: there is nothing to download and nothing to install — run it straight from the directory where you extracted the archive:

tar -xzf filename.tar.gz
cd blurit-op

Replace "filename" with the actual name of the archive that was provided to you.

At the end of the requirements run, the command detects that blurit is not in your PATH and offers to copy it there for you, so you don't have to do it by hand.

Step 2 : run ./blurit requirements

If you are installing BlurIt on an old server and want to maintain specific version dependencies, we don't recommend to use it but rather to manually install everything.

chmod +x ./blurit
./blurit requirements

Run the command as your regular user, not with sudo. It calls sudo on its own for every step that needs root privileges (you will be prompted for your password), and it adds the user running it to the docker group. Launching it with sudo would add root to that group instead, and you would still not be able to use Docker as your own user afterwards.

The command is interactive and safe to re-run: it only checks what is already present and installs what is missing, so you can run it again after fixing a problem or after adding a GPU to the server.

1. Distribution detection

The distribution and version are read from /etc/os-release (or lsb_release).

An unsupported distribution or version does not stop the command: it is reported as a failed OS requirement and the other steps still run, but the installation is then unsupported and may fail.

2. Installation type

You are asked to choose between a CPU-only and a GPU installation:

Choose installation type:
1) CPU-only installation (no NVIDIA components)
2) GPU installation (with NVIDIA drivers and container toolkit)
Enter your choice (1 or 2):

Choose 2 if the server has an NVIDIA GPU — this is the recommended setup for Blurit OP, and by far the fastest. Choose 1 for a CPU-only server: all NVIDIA-related steps are then skipped.

3. GPU detection

For a GPU installation, the GPU is detected with lspci (or nvidia-smi when lspci is not available) and the driver to install is selected accordingly — a Tesla (datacenter) GPU uses the NVIDIA CUDA repositories, a standard GPU uses the distribution's driver packages.

If no NVIDIA GPU is found, the NVIDIA components are skipped and a warning is printed at the end of the run.

4. Requirements report

The status of every requirement is then printed, in green when satisfied, in red when missing, in yellow for warnings:

Requirements:
  OS............................. Ubuntu 22.04.4 LTS
  RAM (>= 16 GB)................. Installed
  Disk space (>= 60 GB).......... Yes
  Docker......................... Not installed
  Docker Compose (optional)...... Not installed
  User in docker group (optional) Missing
  Shell completion (optional).... Not installed
  NVIDIA driver.................. Not installed
  NVIDIA Container Toolkit....... Not installed
  Docker NVIDIA runtime.......... Not configured

If you run blurit requirmeents after havign already installed Blurit OP, the CLI may report insufficient disk space. In that cas, you can probably safely ignore the warning as no extra installation files will be downloaded.

Requirement	What is checked	What is installed / configured
OS	Distribution and version are in the supported list	Nothing — informational warning only
RAM (>= 16 GB)	Total RAM reported by /proc/meminfo	Nothing — a warning is printed below 16 GB, BlurIt may not run optimally
Disk space (>= 60 GB)	Free space on the Docker data directory (docker info -f '{{.DockerRootDir}}', usually /var/lib/docker)	Nothing — a warning is printed below 60 GB, BlurIt may not run optimally
Docker	The docker command exists	Docker CE from Docker's official repository (docker-ce, docker-ce-cli, containerd.io, buildx and compose plugins)
Docker Compose	docker compose version succeeds, or docker-compose exists	Installed with Docker; on Amazon Linux, the latest Compose plugin is downloaded into /usr/local/lib/docker/cli-plugins
User in docker group	Your user belongs to the docker group	sudo usermod -aG docker $USER, so that BlurIt can be operated without sudo
Shell completion	The completion line is present in your ~/.bashrc or ~/.zshrc	Adds the blurit auto-completion line to that file
NVIDIA driver	The nvidia-smi command exists	The NVIDIA driver for your distribution and GPU model (GPU installation only)
NVIDIA Container Toolkit	The nvidia-ctk command exists	nvidia-container-toolkit, then nvidia-ctk runtime configure --runtime=docker and a Docker restart (GPU installation only)
Docker NVIDIA runtime	"default-runtime": "nvidia" is set in /etc/docker/daemon.json	Writes that runtime configuration into /etc/docker/daemon.json and restarts Docker (GPU installation only)

Requirements marked (optional) are not blocking: BlurIt runs without them, but you will have to prefix every Docker command with sudo if you skip the docker group.

5. Selecting what to install

Only the missing requirements are then offered, in a checkbox list where everything is selected by default:

Select requirements to install (up/down to move, space to toggle, enter to confirm):
> [x] Docker
  [x] Docker Compose
  [x] User in docker group
  [x] Shell completion
  [x] NVIDIA driver
  [x] NVIDIA Container Toolkit
  [x] Docker NVIDIA runtime

• up / down arrows move the cursor,

• space toggles the entry under the cursor,

• enter confirms and starts the installation.

Uncheck an entry to keep an existing setup of your own — for example a Docker installation managed by your IT department. If you uncheck everything, nothing is installed.

If all requirements are already satisfied, the command simply prints All requirements are already installed. and moves on to the final steps.

6. Installation

Each selected requirement is installed in turn, then re-checked. The output of the underlying package manager is displayed as it runs, and sudo may ask for your password. If a requirement is still missing after its installation step, the command stops with an error so that you can fix the cause before going further.

Installing an NVIDIA driver takes several minutes, during which the server may seem unresponsive. This is expected — do not interrupt the command.

7. Final steps

Depending on what was installed, the command ends with up to three questions:

1. Reboot — if an NVIDIA driver was installed, a reboot is required to load it:

   NVIDIA drivers were installed and require a reboot to load. Reboot now? THE SERVER WILL BE RESTARTED! [y/N]:

   Answer y to reboot immediately, or N to reboot yourself later. The driver is not usable until the server has rebooted.

2. newgrp docker — if your user was added to the docker group, the change only applies to new sessions. Accepting opens a new shell with the group applied, so you can continue right away; otherwise, log out and log back in.

3. Installing the CLI — if blurit is not in your PATH, the command offers to copy the binary there:

   The 'blurit' binary is not in your PATH, do you want to install it somewhere in your PATH?
   Be sure to choose a directory that is listed in your PATH
   Installation directory [/usr/local/sbin]:

   Accept the proposed directory (or type another one that is in your PATH) and every later command can be called as blurit ... instead of ./blurit ....

If you would rather install it yourself, move it to a directory of your PATH and drop the ./ prefix from every command below:

sudo mv blurit /usr/local/sbin

And if running ./blurit reports Permission denied, the executable bit was lost while copying the archive around: restore it with chmod +x blurit.

Amazon Linux

Amazon Linux images are expected to ship with Docker and, for GPU instances, with the NVIDIA driver, the container toolkit and the Docker NVIDIA runtime already installed and configured. The command therefore only checks them, and installs Docker Compose when it is missing. A missing docker binary is a fatal error: install Docker before running blurit requirements.

Verifying the installation

After the run (and after the reboot, if there was one):

docker version            # Docker is installed and usable without sudo
docker compose version    # Docker Compose plugin
nvidia-smi                # GPU installation only: driver loaded, GPU visible

For a GPU installation, you can also check that Docker can reach the GPU:

docker run --rm --gpus all nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi

Troubleshooting

• permission denied on the Docker socket — your session does not have the docker group yet. Run newgrp docker, or log out and log back in.

• blurit: command not found — the binary is not in your PATH. Call it from the directory where you extracted the archive with ./blurit, or move it to a directory of your PATH (sudo mv blurit /usr/local/sbin).

• A requirement fails to install — the command stops with <requirement> installation failed. Fix the cause reported by the package manager above that line (no network access, proxy, repository not reachable, disk full), then run ./blurit requirements again: the steps already completed are skipped.

• Unsupported distribution — the OS requirement is red. You can continue at your own risk, or follow the manual installation.

Running it as part of the installation

The requirements steps can also be run right before the BlurIt configuration, in a single command:

./blurit install --requirements

🚨 If NVIDIA drivers were installed, please reboot the server to load them, then go to the Configuration page. 🚨

Choice 2 : manual dependencies installation

Nvidia drivers Installation

1. Download the NVIDIA driver >= 470 runfile from: https://www.nvidia.com/Download/index.aspx?lang=en-us

2. Open a terminal and navigate to the directory where the downloaded runfile is located.

3. $ chmod +x <filename>.run Replace <filename> with the actual name of the downloaded runfile.

4. $ sudo ./<filename>.run

5. $ sudo reboot

Docker Installation

The Docker installation documentation is here.

Ubuntu

sudo apt-get update
sudo apt-get install -y apt-transport-https ca-certificates curl gnupg-agent software-properties-common
sudo install -m 0755 -d /etc/apt/keyrings

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
sudo chmod a+r /etc/apt/keyrings/docker.gpg
echo "deb [arch="$(dpkg --print-architecture)" signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu "$(. /etc/os-release && echo "$VERSION_CODENAME")" stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

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

Debian

sudo apt-get update
sudo apt-get install ca-certificates curl gnupg
sudo install -m 0755 -d /etc/apt/keyrings

curl -fsSL https://download.docker.com/linux/debian/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
sudo chmod a+r /etc/apt/keyrings/docker.gpg
echo "deb [arch="$(dpkg --print-architecture)" signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/debian "$(. /etc/os-release && echo "$VERSION_CODENAME")" stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

sudo apt-get update
sudo apt-get install docker-ce docker-ce-cli containerd.io docker-buildx-plugin

Centos

sudo yum install -y yum-utils
sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
sudo yum install docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin
sudo systemctl start docker

RHEL

sudo yum install -y yum-utils
sudo yum-config-manager --add-repo https://download.docker.com/linux/rhel/docker-ce.repo
sudo yum install docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin
sudo systemctl start docker

Post-install docker linux (required):

sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker

Nvidia-docker installation

Nvidia-docker installation documentation is here.

Ubuntu / Debian

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit

Centos / RHEL

curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo | \
  sudo tee /etc/yum.repos.d/nvidia-container-toolkit.repo

sudo yum clean expire-cache
sudo yum install -y nvidia-container-toolkit

Configure the Docker daemon to recognize the NVIDIA Container Runtime:

sudo nvidia-ctk runtime configure --runtime=docker

Restart the Docker daemon to complete the installation after setting the default runtime:

sudo systemctl restart docker

Test if Nvidia-docker is installed:

sudo docker run --rm --runtime=nvidia --gpus all nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi

This should result in a console output shown below:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.51.06    Driver Version: 470.51.06    CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            On   | 00000000:00:1E.0 Off |                    0 |
| N/A   34C    P8     9W /  70W |      0MiB / 15109MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

Configuring daemon.json

Open /etc/docker/daemon.json with your favorite text editor.

Add "default-runtime": "nvidia", to the json. Exemple:

/etc/docker/daemon.json

{
  "default-runtime": "nvidia",
  "runtimes": {
    "nvidia": {
      "args": [],
      "path": "nvidia-container-runtime"
    }
  }
}