Docker, singularity, apptainer
Table of Contents
1. Docker
1.1. Docker setup
1.1.1. Execute docker command with non-root user
Restart the system after executing the following commands.
sudo groupadd docker sudo usermod -aG docker $USER
1.2. Important docker commands
Typically docker needs sudo access.
1.3. Docker list commands
| Command | Description |
|---|---|
| docker image list | List docker images in current system. |
| docker ps | Lists running containers. (ps = process status) |
| docker ps -a | Lists running and excited containers. |
1.4. Starting containers
| Command | Description |
|---|---|
| docker container run -it <image>:<tag> | Starting image in iteractive mode |
| docker container run -it –runtime=nvidia –gpus all <image>:<tag> | Above + accessing nvidia GPUs |
| docker start <container id> | Starting stopped container |
| docker container run -v <host path>:<docker path> -it | o |
1.5. Attaching
| Command | Description |
|---|---|
| docker attach <container id> | Attaching to running container. |
| docker exec -it <container id> bash | Attaching another terminal to running container. |
1.6. Removing
| Command | Description |
|---|---|
| docker rmi <image id> | removes docker images |
| docker rm <container id> | removes stopped container |
| docker system prune | Removes all stopped containers |
2. Apptainer
2.1. Converting docker to singularity
2.1.1. Pull + convert
apptainer pull <sif file> <docker hub link> # Example: apptainer pull cuda_12.6.3.sif docker://nvidia/cuda:12.6.3-cudnn-devel-ubuntu22.04
2.1.2. Convert local image
This involves (1) converting to tar file and (2) converting to sif file.
# First convert to tar file docker save -o <name>.tar <image> docker save -o mlev_39_tf260.tar 708d3b89ade9 # Build app container apptainer build mlev_39_tf260.sif docker-archive://mlev_39_tf260.tar
2.2. Starting
| Command | Description |
|---|---|
| apptainer shell –nv <sif file> | Start shell with nvidia driver library |