Skip to content

Latest commit

 

History

History
539 lines (374 loc) · 20 KB

useful_tips.md

File metadata and controls

539 lines (374 loc) · 20 KB
Raw

Useful tips

Launch a distrobox from you applications list

Starting from distrobox 1.4.0, containers created will automatically generate a desktop entry. For containers generated with older versions, you can use:

distrobox generate-entry you-container-name

To delete it:

distrobox generate-entry you-container-name --delete

Create a distrobox with a custom HOME directory

distrobox create supports the use of the --home flag, as specified in the usage HERE

Simply use:

distrobox create --name test --image your-chosen-image:tag --home /your/custom/home

Mount additional volumes in a distrobox

distrobox create supports the use of the --volume flag, as specified in the usage HERE

Simply use:

distrobox create --name test --image your-chosen-image:tag --volume /your/custom/volume/path

Use a different shell than the host

From version 1.4.0, distrobox enter will execute the login shell of the container's user by default. So, just change the default shell in the container using:

chsh -s /bin/shell-to-use

exit and log back in the container.

For version older than 1.4.0, distrobox will pick up the shell from the host and use it inside the container. If you want a different one you can use:

SHELL=/bin/zsh distrobox create -n test
SHELL=/bin/zsh distrobox enter test

Run the container with real root

When using podman, distrobox will prefer to use rootless containers. In this mode the root user inside the container is not the real root user of the host. But it still has the same privileges as your normal $USER.

But what if you really really need those root privileges even inside the container?

Running sudo distrobox is not supported, instead, it is better to simply use normal command with the --root or -r flag, so that distrobox can still integrate better with your $USER.

:~$ distrobox create --name test --image your-chosen-image:tag --root

Another use case, what if you want or need to run distrobox with the root user, in a login shell?

Before the 1.4.3 release, it wasn't possible. We couldn't make a distinction between someone running distrobox vi sudo from someone logged in as the root user in a shell. Now things are as easy as it would be if you were creating a rootless container:

:~# distrobox create --name your-container --pull --image your-chosen-image:tag`

And:

:~# distrobox enter your-container`

We trust you already know the implications of running distrobox, as well as anything else, with the root user and that with great power comes great responsibilities.

Run Debian/Ubuntu container behind proxy

It might be that you're trying to set-up your distrobox, but you're stuck behind a proxy. A simple solution can be crafted using pre-init-hooks

proxy=http://my_proxy.domain.example:3128
t="echo 'Acquire::http::Proxy \\\""${proxy}"\\\";' > /etc/apt/apt.conf.d/proxy.conf; echo 'Acquire::https::Proxy \\\""${proxy}"\\\";' >> /etc/apt/apt.conf.d/proxy.conf;"
http_proxy="${proxy}" distrobox create --image debian --name deb --pre-init-hooks "${t}"

This way, we're configuring apt before using it.

Using a command other than sudo to run a rootful container

When using the --root option with Distrobox, internally, it uses sudo to be able to interact with the rootful container through podman/docker, which will prompt for a valid root password on the terminal. However, some users might prefer to use a command other than sudo in order to authenticate as root; for example, pkexec could be used to display a graphical authentication prompt. If you need this, make sure to specify the desired command through the DBX_SUDO_PROGRAM environment variable (supported by most distrobox subcommands), alongside --root. Sample usage:

DBX_SUDO_PROGRAM="pkexec" distrobox create --name test --image your-chosen-image:tag --root

Additionally, you may also have any further distrobox commands use pkexec (for example) for rootful containers by appending the line distrobox_sudo_program="pkexec" (replace pkexec with the desired program) to one of the config file paths that distrobox supports; for example, to '~/.distroboxrc'.

It is also worth noting that, if your sudo program does not have persistence (i.e., cooldown before asking for the root password again after a successful authentication) configured, then you may have to enter the root password multiple times, as distrobox calls multiple podman/docker commands under the hood. In order to avoid this, it is recommended to either configure your sudo program to be persistent, or, if that's not feasible, use sudo whenever possible (which has persistence enabled by default).

However, if you'd like to have a graphical authentication prompt, but would also like to benefit from sudo's persistence (to avoid prompting for a password multiple times in a row), you may specify sudo --askpass as the sudo program. The --askpass option makes sudo launch the program in the path (or name, if it is in $PATH) specified by the SUDO_ASKPASS environment variable, and uses its output (to stdout) as the password input to authenticate as root. If unsuccessful, it launches the program again, until either it outputs the correct password, the user cancels the operation, or a limit of amount of authentication attempts is reached.

So, for example, assume you'd like to use zenity --password to prompt for the sudo password. You may save a script, e.g. my-password-prompt, to somewhere in your machine - say, to ~/.local/bin/my-password-prompt - with the following contents:

#!/bin/sh
zenity --password

Make it executable using, for example, chmod (in the example, by running chmod +x ~/.local/bin/my-password-prompt - replace with the path to your script). Afterwards, make sure SUDO_ASKPASS is set to your newly-created script's path, and also ensure DBX_SUDO_PROGRAM is set to sudo --askpass, and you should be good to go. For example, running the below command should only prompt the root authentication GUI once throughout the whole process:

SUDO_ASKPASS="$HOME/.local/bin/my-password-prompt" DBX_SUDO_PROGRAM="sudo --askpass" distrobox-ephemeral -r

You may make these options persist by specifying those environment variables in your shell's rc file (such as ~/.bashrc). Note that this will also work if distrobox_sudo_program="sudo --askpass" is specified in one of distrobox's config files (such as ~/.distroboxrc), alongside export SUDO_ASKPASS="/path/to/password/prompt/program" (for example - however, this last line is usually better suited to your shell's rc file).

Duplicate an existing distrobox

It can be useful to just duplicate an already set up environment, to do this, distrobox create supports the use of the --clone flag, as specified in the usage HERE

Simply use:

distrobox create --name test --clone name-of-distrobox-to-clone

Export to the host

Distrobox supports exporting to the host either binaries, applications or systemd services. Head over the usage page to have an explanation and examples.

Execute commands on the host

You can check this little post about executing commands on the host.

Enable SSH X-Forwarding when SSH-ing in a distrobox

SSH X-forwarding by default will not work because the container hostname is different from the host's one. You can create a distrobox with will have the same hostname as the host by creating it with the following init-hook:

distrobox create --name test --image your-chosen-image:tag \
                  --init-hooks '"$(uname -n)" > /etc/hostname'`

This will ensure SSH X-Forwarding will work when SSH-ing inside the distrobox:

ssh -X myhost distrobox enter test -- xclock

Using podman or docker inside a distrobox

You can easily control host's instance of docker or podman, using distrobox-host-exec You can use:

sudo ln -s /usr/bin/distrobox-host-exec /usr/local/bin/podman

or

sudo ln -s /usr/bin/distrobox-host-exec /usr/local/bin/docker

This will create a podman or docker command inside the distrobox that will transparently execute the command on the host.

Using init system inside a distrobox

You can use an init system inside the container. You can either use supported pre-created images, or have to add additional packages.

Example of such images are:

  • docker.io/almalinux/8-init
  • registry.access.redhat.com/ubi7/ubi-init
  • registry.access.redhat.com/ubi8/ubi-init
  • registry.access.redhat.com/ubi9/ubi-init
  • registry.opensuse.org/opensuse/leap:latest
  • registry.opensuse.org/opensuse/tumbleweed:latest

You can use such feature using:

distrobox create -i docker.io/almalinux/8-init --init --name test

If you want to use a non-pre-create image, you'll need to add the additional package:

distrobox create -i debian:unstable --init --additional-packages "systemd" -n test distrobox create -i ubuntu:18.04 --init --additional-packages "systemd" -n test distrobox create -i registry.fedoraproject.org/fedora:38 --init --additional-packages "systemd" -n test

Note however that in this mode, you'll not be able to access host's processes from within the container.

Example use:

~$ distrobox create -i docker.io/almalinux/8-init --init --name test

user@test:~$ sudo systemctl enable --now sshd

user@test:~$ sudo systemctl status sshd
    ● sshd.service - OpenSSH server daemon
       Loaded: loaded (sshd.service; enabled; vendor preset: enabled)
       Active: active (running) since Fri 2022-01-28 22:54:50 CET; 17s ago
         Docs: man:sshd(8)
               man:sshd_config(5)
     Main PID: 291 (sshd)

Using distrobox as main cli

In case you want (like me) to use your container as the main CLI environment, it comes handy to use gnome-terminal profiles to create a dedicated setup for it:

Screenshot from 2021-12-19 22-29-08

Personally, I just bind Ctrl-Alt-T to the Distrobox profile and Super+Enter to the Host profile.

For other terminals, there are similar features (profiles) or you can set up a dedicated shortcut to launch a terminal directly in the distrobox

Using a different architecture

In case you want to run a container with a different architecture from your host, you can leverage the use of qemu and support from podman/docker.

Install on your host the following dependencies:

  • qemu
  • qemu-user-static
  • binfmt-support

Then you can easily run the image you like:

~$ uname -m
x86_64
~$ distrobox create -i aarch64/fedora -n fedora-arm64
~$ distrobox enter fedora-arm64
...
user@fedora-arm64:~$ uname -m
aarch64

image

Using the GPU inside the container

For Intel and AMD Gpus, the support is backed in, as the containers will install their latest available mesa/dri drivers.

For NVidia, you can use the --nvidia flag during create, see distrobox-create documentation to discover how to use it.

~$ distrobox create --nvidia --name ubuntu-nvidia --image ubuntu:latest

Using nvidia-container-toolkit

Alternatively from the --nvidia flag, you can use NVidia's own nvidia-container-toolkit. After following the official guide to set nvidia-ctk up you can use it from distrobox doing:

distrobox create --name example-nvidia-toolkit --additional-flags "--runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all -e NVIDIA_DRIVER_CAPABILITIES=all" --image nvidia/cuda

Slow creation on podman and image size getting bigger with distrobox create

For rootless podman 3.4.0 and upward, adding this to your ~/.config/containers/storage.conf file will improve container creation speed and fix issues with images getting bigger when using rootless containers.

[storage]
driver = "overlay"

[storage.options.overlay]
mount_program = "/usr/bin/fuse-overlayfs"

Note that this is necessary only on Kernel version older than 5.11 . From version 5.11 onwards native overlayfs is supported and reports noticeable gains in performance as explained HERE

Container save and restore

To save, export and reuse an already configured container, you can leverage podman save or docker save and podman import or docker import to create snapshots of your environment.

To save a container to an image:

with podman:

podman container commit -p distrobox_name image_name_you_choose
podman save image_name_you_choose:latest | gzip > image_name_you_choose.tar.gz

with docker:

docker container commit -p distrobox_name image_name_you_choose
docker save image_name_you_choose:latest | gzip > image_name_you_choose.tar.gz

This will create a tar.gz of the container of your choice at that exact moment.

Now you can backup that archive or transfer it to another host, and to restore it just run

podman load < image_name_you_choose.tar.gz

or

docker load < image_name_you_choose.tar.gz

And create a new container based on that image:

distrobox create --image image_name_you_choose:latest --name distrobox_name
distrobox enter --name distrobox_name

And you're good to go, now you can reproduce your personal environment everywhere in simple (and scriptable) steps.

Check used resources

  • You can always check how much space a distrobox is taking by using podman command:

podman system df -v or docker system df -v

Pre-installing additional package repositories

On Red Hat Enterprise Linux and its derivatives, the amount of packages in the base repositories is limited, and additional packages need to be brought in by enabling additional repositories such as EPEL.

You can use --init-hooks to automate this, but this does not solve the issue for package installations done during initialization itself, e.g. if the shell you use on the host is not available in the default repos (e.g. fish).

Use the pre-initialization hooks for this:

distrobox create -i docker.io/almalinux/8-init --init --name test --pre-init-hooks "dnf -y install dnf-plugins-core && dnf config-manager --enable powertools && dnf -y install epel-release"
distrobox create -i docker.io/library/almalinux:9 -n alma9 --pre-init-hooks "dnf -y install dnf-plugins-core && dnf config-manager --enable crb && dnf -y install epel-release"
distrobox create -i quay.io/centos/centos:stream9 c9s --pre-init-hooks "dnf -y install dnf-plugins-core && dnf config-manager --enable crb && dnf -y install epel-next-release"

Apply resource limitation on the fly

Podman has --cpuset-cpus and --memory flags to apply limitation on how much resources a container can use. However, these flags only work during container creation (podman create / podman run) and not after it's created (podman exec, which is used by Distrobox to execute commands inside of container), which means changing resource limitation requires recreation of a container.

Nonetheless you can still apply resource limitation using systemd's resource control functionality. It's not recommended to pass resource limitation arguments (e.g. --cpuset-cpus and --memory) to distrobox create --additional-flags as systemd already provides much more flexible resource control functionality.

To list all distroboxes and their full IDs:

podman ps --all --no-trunc --format "{{.Names}} {{.ID}} {{.Labels}}" | grep "manager:distrobox" | cut -d " " -f1,2 | column -t
  • Removing --all flag will cause the output to only contain currently running distroboxes

To check your container status with systemctl:

systemctl --user status libpod-$UUID.scope

  • Your distrobox needs to be running for its scope to present (e.g. distrobox enter before running this command)

  • Replace $UUID with your container's real full ID

  • To make things easier when tweaking properties, optionally set a environment variable for the current shell:

    bash/zsh:

    UUID=XXXXXXXXX

    fish:

    set UUID XXXXXXXXX

Everything provided by systemd.resource-control could be applied to your distrobox. For example:

To make your distrobox only run on CPU0 and CPU1:

systemctl --user set-property libpod-$UUID.scope AllowedCPUs=0,1

To hard throttle your distrobox to not use above 20% of CPU:

systemctl --user set-property libpod-$UUID.scope CPUQuota=20%

To limit your distrobox's maximum amount of memory:

systemctl --user set-property libpod-$UUID.scope MemoryMax=2G

To give your distrobox less IO bandwidth when IO is overloaded:

systemctl --user set-property libpod-$UUID.scope IOWeight=1
  • IOWeight accepts value from 1 to 10000, higher means more bandwidth.

To see all applicable properties:

man systemd.resource-control

Changes are transient, meaning you lose the resource limitation properties when distrobox is stopped and restarted.

To make certain changes persistent, first check the currently active properties:

systemctl --user status libpod-$UUID.scope

Look for the Drop-In lines. Something like this should be shown:

    Drop-In: /run/user/1000/systemd/transient/libpod-45ae38d61c9a636230b2ba89ea07792d662e01cd9ee38d04feb0a994b039a271.scope.d
             └─50-AllowedCPUs.conf

Move the transient overrides to persistent overrides:

mkdir -p ~/.config/systemd/user/libpod-$UUID.scope.d
mv --target-directory="$HOME/.config/systemd/user/libpod-$UUID.scope.d" \
  "/run/user/$(id -u)/systemd/transient/libpod-$UUID.scope.d/50-AllowedCPUs.conf"
  • Replace $(id -u) with your real user id if it did not get expanded properly.
  • 50-AllowedCPUs.conf is only an example. Replace it with something you want to keep persistently.

Then reload systemd daemon to apply the changes:

systemctl --user daemon-reload