Docker Tips

Follow this to get a good command of Docker usage.

Basic Commands

The Reference documentation | Docker Docs gives the most complete guidance.

But here, only swift commnads listed which are sufficient for general using.

BTW I documented how to run a mysql docker container in First how-to-do note, including initial steps.

Run the Container

Docker can be used as microservice, while for non-professional developers, we use it as lightweight virtual machine.

$ docker run -itd image_name /bin/bash
# "-itd" means to run interactive /bin/bash, on tty, in background
#  which makes container stay alive and wait for attaching.  
# can also add "--name container_name" to name it, or "-p host_port:guest_port" to set port forwarding

# start the contaier
$ docker start container_name
# "attach" to it, while not `docker attach`
$ docker exec -it container_name /bin/bash
# stop the container
$ docker stop container_name

# attention, if you attach to container...
$ docker attach container_name
# dont `exit` cuz it will stop the container, try "Ctrl+P+Q"

Migrate Images and Containers

To migrate images, use save and load.
First save the image:

$ docker save -o image.tar image_id_or_image_name
# or
$ docker save image_id_or_image_name > image.tar
# to pack many images together
$ docker save docker save -o images.tar image_1 image_2

Then load the image:

$ docker load -i image.tar
# or 
$ docker load < image.tar

New image name stays the same with the old one.

To migrate containers, use export and import.
First export the container:

$ docker export container_id_or_container_name > container.tar
# or 
$ docker export --output="container.tar" container_id_or_container_name

Then import the container:

$ docker import - new_container_name < container.tar

You see, you can name the new image this step.

Get GPU support from Nvidia

Check this out: Installing the NVIDIA Container Toolkit for updated installation.

First to figure out the differences between: nvidia-docker, nvidia-docker2 and nvidia-container-toolkits.

The nvidia-docker and nvidia-docker2 are archived(Dec 6, 2023) old repos.
In nvidia-docker age it can be runned by

$ nvidia-docker run XXX

In nvidia-docker2 the above nvidia-docker command is equal to:

$ docker run --runtime=nvidia XXX 

The relationship between these methods:

Backward compatibility To help transitioning code from 1.0 to 2.0, a bash script is provided in /usr/bin/nvidia-docker for backward compatibility. It will automatically inject the --runtime=nvidia argument and convert NV_GPU to NVIDIA_VISIBLE_DEVICES.

Up to now, NVIDIA Container Toolkit is the latest solution, as the sample usage goes like:

$ sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

Then output panel is like

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 535.86.10    Driver Version: 535.86.10    CUDA Version: 12.2     |
|-------------------------------+----------------------+----------------------+
| 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 |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

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

The --runtime=nvidia can be omitted, while the --gpus all is essential.
The ubuntu image here can be replaced with images listed on Nvidia CUDA Linux Container Image Supported tags.

Then process as the doc goes.