Difference between revisions of "LXD"

From ArchWiki
Jump to navigation Jump to search
(This is explained in the section above. We don't duplicate information. Undo revision 597113 by Thomas145 (talk))
Tag: Undo
(→‎No IPv4 with systemd-networkd: - reword, fix formatting and add a bit more info about the issue)
Line 118: Line 118:
  
  
systemd version 244.1 detects if {{ic|/sys}} is writeable by containers. If it is, udev is automatically started and breaks IPv4 in unprivileged containers. See [https://github.com/systemd/systemd-stable/commit/96d7083c5499b264ecebd6a30a92e0e8fda14cd5 commit bf331d8] and [https://discuss.linuxcontainers.org/t/no-ipv4-on-arch-linux-containers/6395 discussion on linuxcontainers].
+
Starting with version version 244.1, systemd detects if {{ic|/sys}} is writable by containers. If it is, udev is automatically started and breaks IPv4 in unprivileged containers. See [https://github.com/systemd/systemd-stable/commit/96d7083c5499b264ecebd6a30a92e0e8fda14cd5 commit bf331d8] and [https://discuss.linuxcontainers.org/t/no-ipv4-on-arch-linux-containers/6395 discussion on linuxcontainers].
  
To work around this two ways;
+
On containers created past 2020, there should already be a {{ic|systemd.networkd.service}} override to work around this issue, create it if it is not:
  
 
Add {{ic|1=raw.lxc: lxc.mount.auto = proc:rw sys:ro}} in the profile to the container to ensure {{ic|/sys}} is read-only.
 
 
Or, override {{ic|systemd.networkd.service}};
 
 
{{hc|1=/etc/systemd/system/systemd-networkd.service.d/lxc.conf|2=
 
{{hc|1=/etc/systemd/system/systemd-networkd.service.d/lxc.conf|2=
 
[Service]
 
[Service]
 
BindReadOnlyPaths=/sys
 
BindReadOnlyPaths=/sys
 
}}
 
}}
 +
 +
You could also work around this issue by setting {{ic|1=raw.lxc: lxc.mount.auto = proc:rw sys:ro}} in the profile of the container to ensure {{ic|/sys}} is read-only for the entire container, although this may be problematic, as per the linked discussion above.
  
 
== Uninstall ==
 
== Uninstall ==

Revision as of 19:54, 9 February 2020

LXD is a container and virtual-machine "hypervisor" and a new user experience for Linux Containers.

Setup

Required software

Install LXC and the lxd package, then start lxd.service.

See Linux Containers#Enable support to run unprivileged containers (optional) if you want to run unprivileged containers. Otherwise see #Launching container without CONFIG_USER_NS.

Alternative installation method

An alternative method of installation is via snapd, by installing the snapdAUR package and running:

# snap install lxd

Configure LXD

LXD needs to configure a storage pool, and (if you want internet access) network configuration. To do so, run the following as root:

# lxd init

Accessing LXD as a unprivileged user

By default the LXD daemon allows users in the lxd group access, so add your user to the group:

# usermod -a -G lxd <user>
Warning: Anyone added to the lxd group is root equivalent. More information here and here.

Usage

Create container

LXD has two parts, the daemon (the lxd binary), and the client (the lxc binary). Now that the daemon is all configured and running, you can create a container:

$ lxc launch ubuntu:16.04

Alternatively, you can also use a remote LXD host as a source of images. One comes pre-configured in LXD, called "images" (images.linuxcontainers.org)

$ lxc launch images:centos/7/amd64 centos

To create an amd64 Arch container:

$ lxc launch images:archlinux/current/amd64 arch

LXD Networking

LXD uses LXC's networking capabilities. By default it connects containers to the lxcbr0 network device. Refer to the LXC documentation on network configuration to set up a bridge for your containers.

If you want to use a different interface than lxcbr0 edit the default using the lxc command line tool:

$ lxc profile edit default

An editor will open with a config file that by default contains:

name: default
config: {}
devices:
  eth0:
    name: eth0
    nictype: bridged
    parent: lxcbr0
    type: nic

You can set the parent parameter to whichever bridge you want LXD to attach the containers to by default.

Troubleshooting

Check kernel config

By default Arch Linux kernel is compiled correctly for Linux Containers and its frontend LXD. However, if you're using a custom kernel, or changed kernel options the kernel might be configured incorrectly. Verify that the running kernel is properly configured to run a container:

$ lxc-checkconfig

Launching container without CONFIG_USER_NS

For launching images you must provide security.privileged=true during image creation:

$ lxc launch ubuntu:16.04 ubuntu -c security.privileged=true

Or for already existed image you may edit config:

$ lxc config edit ubuntu
name: ubuntu
profiles:
- default
config:
  ...
  security.privileged: "true"
  ...
devices:
  root:
    path: /
    type: disk
ephemeral: false

Or to enable security.privileged=true for new containers, edit the config for the default profile:

$ lxc profile edit default

Resource limits are not applied when viewed from inside a container

The service lxcfs (found in the Community repository) needs to be installed and started :

$ systemctl start lxcfs

lxd will need to be restarted. Enable lxcfs for the service to be started at boot time.

Starting a VM fails

Arch Linux does not distribute secure boot signed ovmf firmware, to boot VMs you need to disable secure boot for the time being.

$ lxc launch ubuntu:18.04 test-vm --vm -c security.secureboot=false

This can also be added to the default profile.

No IPv4 with systemd-networkd

Starting with version version 244.1, systemd detects if /sys is writable by containers. If it is, udev is automatically started and breaks IPv4 in unprivileged containers. See commit bf331d8 and discussion on linuxcontainers.

On containers created past 2020, there should already be a systemd.networkd.service override to work around this issue, create it if it is not:

/etc/systemd/system/systemd-networkd.service.d/lxc.conf
[Service]
BindReadOnlyPaths=/sys

You could also work around this issue by setting raw.lxc: lxc.mount.auto = proc:rw sys:ro in the profile of the container to ensure /sys is read-only for the entire container, although this may be problematic, as per the linked discussion above.

Uninstall

Stop and disable lxd.service and lxd.socket:

Then uninstall the package via pacman:

 # pacman -R lxd

If you uninstalled the package without disabling the service, you might have a lingering broken symlink at /etc/systemd/system/multi-user.wants/lxd.service.

If you want to remove all data:

 # rm -r /var/lib/lxd

If you used any of the example networking configs, you should remove those as well.

See also