Difference between revisions of "Linux Containers"

From ArchWiki
Jump to: navigation, search
m (/etc/rc.sysinit replacement: route to ip route.)
m (Error: Failed to initialize cgroup driver / PAM unable to open pam_cgfs.so: typo)
 
(294 intermediate revisions by 73 users not shown)
Line 1: Line 1:
 
[[Category:Security]]
 
[[Category:Security]]
 
[[Category:Virtualization]]
 
[[Category:Virtualization]]
{{Stub|Currently just a rough draft... I think I will need to restructure this a bit and I have also noticed I have become a bit too verbose -_-;; I will be along shortly to complete this as well as clean it up.}}
+
[[ja:Linux Containers]]
 +
[[pt:Linux Containers]]
 +
{{Related articles start}}
 +
{{Related|AirVPN}}
 +
{{Related|ABS}}
 +
{{Related|Cgroups}}
 +
{{Related|Docker}}
 +
{{Related|LXD}}
 +
{{Related|OpenVPN}}
 +
{{Related|OpenVPN (client) in Linux containers}}
 +
{{Related|OpenVPN (server) in Linux containers}}
 +
{{Related|PeerGuardian Linux}}
 +
{{Related|systemd-nspawn}}
 +
{{Related articles end}}
  
==Introduction==
+
[https://linuxcontainers.org/ Linux Containers] (LXC) is an operating-system-level virtualization method for running multiple isolated Linux systems (containers) on a single control host (LXC host). It does not provide a virtual machine, but rather provides a virtual environment that has its own CPU, memory, block I/O, network, etc. space and the resource control mechanism. This is provided by  [[Wikipedia:Linux namespaces|namespaces]] and [[cgroups]] features in Linux kernel on LXC host. It is similar to a chroot, but offers much more isolation.
  
===Synopsis===
+
Alternatives for using containers are [[systemd-nspawn]], [[docker]] or {{Pkg|rkt}}.
  
Linux Containers (LXC) are an operating system-level virtualization method for running multiple isolated server installs (containers) on a single control host. LXC does not provide a virtual machine, but rather provides a virtual environment that has its own process and network space. It is similar to a chroot, but offers much more isolation.
+
== Privileged containers or unprivileged containers ==
 +
LXCs can be setup to run in either ''privileged'' or ''unprivileged'' configurations.
  
===About this HowTo===
+
In general, running an ''unprivileged'' container is [https://www.stgraber.org/2014/01/17/lxc-1-0-unprivileged-containers considered safer] than running a ''privileged'' container since ''unprivileged'' containers have an increased degree of isolation by virtue of their design.  Key to this is the mapping of the root UID in the container to a non-root UID on the host which makes it more difficult for a hack within the container to lead to consequences on host system.  In other words, if an attacker manages to escape the container, he or she should find themselves with no rights on the host.
  
This document is intended as an overview on setting up and deploying containers, and is not an in depth detailed instruction by instruction guide. A certain amount of prerequisite knowledge and skills are assumed (running commands as root, kernel configuration, mounting filesystems, shell scripting, chroot type environments, networking setup, etc).
+
The Arch packages currently provide out-of-the-box support for ''privileged'' containers. ''Unprivileged'' containers are only available for the system  administrator with additional kernel configuration. This is due to the current Arch {{pkg|linux}} kernel shipping with user namespaces disabled for normal users.  This article contains information for users to run either type of container, but additional setup is required to use ''unprivileged'' containers.
  
Much of this was taken verbatim from [http://lxc.teegra.net/ Dwight Schauer], [http://tuxce.selfip.org/informatique/conteneurs-linux-lxc Tuxce] and [http://artisan.karma-lab.net/node/1749 Ulhume]. It has been copied here both to enable to community to share their collective wisdom and to expand on a few points.
+
=== An example to illustrate unprivileged containers ===
  
===Less verbose tutorial===
+
To illustrate the power of UID mapping, consider the output below from a running, ''unprivileged'' container.  Therein, we see the containerized processes owned by the containerized root user in the output of {{ic|ps}}:
  
[[User:Delerious010|Delerious010]] 21:43, 1 December 2009 (EST) I have come to realize I have added a lot of text to this HowTo. If you would like something more streamlined, please head on over to [http://lxc.teegra.net/ http://lxc.teegra.net/] for Dwight's excellent guide.
+
[root@unprivileged_container /]# ps -ef | head -n 5
 +
UID        PID  PPID  C STIME TTY          TIME CMD
 +
root        1    0  0 17:49 ?        00:00:00 /sbin/init
 +
root        14    1  0 17:49 ?        00:00:00 /usr/lib/systemd/systemd-journald
 +
dbus        25    1 0 17:49 ?        00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation
 +
systemd+    26    1  0 17:49 ?        00:00:00 /usr/lib/systemd/systemd-networkd
  
===Testing capabilities===
+
On the host however, those containerized root processes are running as the mapped user (ID>100000) on the host, not as the root user on the host:
 +
[root@host /]# lxc-info -Ssip --name sandbox
 +
State:          RUNNING
 +
PID:            26204
 +
CPU use:        10.51 seconds
 +
BlkIO use:      244.00 KiB
 +
Memory use:    13.09 MiB
 +
KMem use:      7.21 MiB
  
Once the lxc package is installed, running lxc-checkconfig will print out a list of your system's capabilities
+
[root@host /]# ps -ef | grep 26204 | head -n 5
 +
UID        PID  PPID  C STIME TTY          TIME CMD
 +
100000  26204 26200  0 12:49 ?        00:00:00 /sbin/init
 +
100000  26256 26204  0 12:49 ?        00:00:00 /usr/lib/systemd/systemd-journald
 +
100081  26282 26204  0 12:49 ?        00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation
 +
100000  26284 26204  0 12:49 ?        00:00:00 /usr/lib/systemd/systemd-logind
  
==Host configuration==
+
== Setup ==
 +
=== Required software ===
 +
Installing {{Pkg|lxc}} and {{Pkg|arch-install-scripts}} will allow the host system to run privileged lxcs.
  
===Control group filesystem===
+
==== Enable support to run unprivileged containers (optional) ====
 +
Users wishing to run ''unprivileged'' containers need to complete several additional setup steps.
  
LXC depends on the control group filesystem being mounted. The standard location for it is {{ic|/sys/fs/cgroup}}. If you use systemd, the cgroup filesystem will be mounted automatically, including the default controllers, but with other initsystems you might have to do it yourself:
+
Firstly, a kernel is required that has support for '''User Namespaces''' (a kernel with {{ic|CONFIG_USER_NS}}). All Arch Linux kernels have support for {{ic|CONFIG_USER_NS}}. However, due to more general security concerns, the default Arch kernel does ship with User Namespaces enabled only for the ''root'' user. You have multiple options to create ''unprivileged'' containers:
  
mount -t tmpfs none /sys/fs/cgroup
+
* Start your unprivileged containers only as ''root''.
 +
* Enable the ''sysctl'' setting {{ic|kernel.unprivileged_userns_clone}} to allow normal users to run unprivileged containers. This can be done for the current session with {{ic|1=sysctl kernel.unprivileged_userns_clone=1}} and can be made permanent with {{man|5|sysctl.d}}.
 +
* Enable the cgfs pam module by modifying {{ic|/etc/pam.d/system-login}} to '''additionally''' contain the following line:
 +
session optional pam_cgfs.so -c freezer,memory,name=systemd,unified
  
===Userspace tools===
+
Secondly, modify {{ic|/etc/lxc/default.conf}} to contain the following lines:
 +
lxc.idmap = u 0 100000 65536
 +
lxc.idmap = g 0 100000 65536
  
Install {{Pkg|lxc}} from [community].
+
Finally, create both {{ic|/etc/subuid}} and {{ic|/etc/subgid}} to contain the mapping to the containerized uid/gid pairs for each user who shall be able to run the containers. The example below is simply for the root user (and systemd system unit):
  
===Bridge device setup===
+
{{hc|/etc/subuid|
 +
root:100000:65536
 +
}}
 +
 
 +
{{hc|/etc/subgid|
 +
root:100000:65536
 +
}}
  
The package {{pkg|bridge-utils}}, which provides the {{ic|brctl}} command, provides the ability to set up one or more network bridges to be used by LXC. You can use {{ic|brctl}} directly to set up the bridges, but the process is archaic and probably best avoided.
+
=== Host network configuration ===
 +
LXCs support different virtual network types and devices (see {{man|5|lxc.container.conf}}). A bridge device on the host is required for most types of virtual networking.
  
If using netcfg, you could try to refer to the [[netcfg]] wiki page, but the bridge related documentation has been removed. So when using netcfg, you will be mostly on your own.
+
LXC comes with its own NAT Bridge (lxcbr0).
 +
{{Note|A NAT bridge is a standalone bridge with a private network that is not bridged to the host eth0 or a physical network. It exists as a private subnet in the host.}}
 +
{{Tip|This is quite useful when WIFI is the only option. There have been various attempts of creating Bridges on WIFI without much success.}}
  
OpenVPN has complete instructions for setting up a bridge, so it is probably your best option at the current. [[OpenVPN Bridge]]
+
To use LXC's NAT Bridge you need to create its configuration file:
  
===Starting a container on boot with [[Systemd]]===
+
{{hc|/etc/default/lxc-net|
 +
2=# Leave USE_LXC_BRIDGE as "true" if you want to use lxcbr0 for your
 +
# containers.  Set to "false" if you'll use virbr0 or another existing
 +
# bridge, or mavlan to your host's NIC.
 +
USE_LXC_BRIDGE="true"
  
If you completed a container, starting it when the host boots is possible with the following systemd service template:
+
# If you change the LXC_BRIDGE to something other than lxcbr0, then
 +
# you will also need to update your /etc/lxc/default.conf as well as the
 +
# configuration (/var/lib/lxc/<container>/config) for any containers
 +
# already created using the default config to reflect the new bridge
 +
# name.
 +
# If you have the dnsmasq daemon installed, you'll also have to update
 +
# /etc/dnsmasq.d/lxc and restart the system wide dnsmasq daemon.
 +
LXC_BRIDGE="lxcbr0"
 +
LXC_ADDR="10.0.3.1"
 +
LXC_NETMASK="255.255.255.0"
 +
LXC_NETWORK="10.0.3.0/24"
 +
LXC_DHCP_RANGE="10.0.3.2,10.0.3.254"
 +
LXC_DHCP_MAX="253"
 +
# Uncomment the next line if you'd like to use a conf-file for the lxcbr0
 +
# dnsmasq.  For instance, you can use 'dhcp-host=mail1,10.0.3.100' to have
 +
# container 'mail1' always get ip address 10.0.3.100.
 +
#LXC_DHCP_CONFILE=/etc/lxc/dnsmasq.conf
  
{{bc|1=
+
# Uncomment the next line if you want lxcbr0's dnsmasq to resolve the .lxc
[Unit]
+
# domain.  You can then add "server=/lxc/10.0.3.1' (or your actual $LXC_ADDR)
Description=Linux Container
+
# to your system dnsmasq configuration file (normally /etc/dnsmasq.conf,
After=network.target
+
# or /etc/NetworkManager/dnsmasq.d/lxc.conf on systems that use NetworkManager).
 +
# Once these changes are made, restart the lxc-net and network-manager services.
 +
# 'container1.lxc' will then resolve on your host.
 +
#LXC_DOMAIN="lxc"
 +
}}
  
[Service]
+
{{Tip| Make sure the bridges ip-range does not interfere with your local network.}}
Type=forking
 
ExecStartPre=/bin/mount --make-rprivate /
 
ExecStart=/usr/bin/lxc-start -dn CONTAINER_NAME
 
ExecStop=/usr/bin/lxc-stop -n CONTAINER_NAME
 
  
[Install]
+
Then we need to modify the LXC container template so our containers use our bridge:
WantedBy=multi-user.target
+
 
 +
{{hc|/etc/lxc/default.conf|
 +
2=lxc.net.0.type = veth
 +
lxc.net.0.link = lxcbr0
 +
lxc.net.0.flags = up
 +
lxc.net.0.hwaddr = 00:16:3e:xx:xx:xx
 
}}
 
}}
Replace CONTAINER_NAME with the name of your container and save this file as /etc/systemd/system/lxc-CONTAINER_NAME.service
 
  
==Container setup==
+
You also need to [[install]] {{pkg|dnsmasq}} which is a dependency for lxcbr0.
 +
 
 +
[[Enable]] and/or [[start]] {{ic|lxc-net.service}} to use the bridge:
 +
 
 +
See [[Network bridge]] for more information.
 +
 
 +
=== Alternate Network - Bridge on same network as host ===
  
There are various different means to do this
+
{{Remove|Duplicates [[Bridge with netctl]].}}
  
===Creating the filesystem===
+
If you would like the containers to be on the same network as your host machine and not utilize NAT and firewall forwarding rules you can do the following.
  
====Bootstrap====
+
You will setup a manual bridge using netctl that includes the hosts ethernet adapter. You set your ethernet to not have an ip address and instead assign an ip address to the bridge that it is binded to.  
Bootstrap an install ( [http://blog.mudy.info/tag/mkarchroot/ mkarchroot], [http://wiki.debian.org/Debootstrap debootstrap], [http://www.xen-tools.org/software/rinse/faq.html rinse], [[Install From Existing Linux]] ). You can also just copy/use an existing installation’s complete root filesystem.
 
  
For example, install a small debian to /home/lxc/debianfs
+
In this scenario you will not be using the lxc-net service stop it if you enabled it.
  
  yaourt -S debootstrap # install debootstrap from AUR
+
  # systemctl stop lxc-net
 +
# systemctl disable lxc-net
  
# method 1:
+
{{hc|/etc/netctl/ethernet|
sudo debootstrap wheezy /home/lxc/debianfst http://ftp.us.debian.org/debian  # use us mirror site install wheezy version
+
2=Description="Manual Ethernet setup for bridge"
# or, method 2:  use faster tar ball method
+
Interface=eno1    # your interface
sudo debootstrap --make-tarball wheezy.packages.tgz sid http://debian.osuosl.org/debian/
+
Connection=ethernet
sudo debootstrap --unpack-tarball wheezy.packages.tgz wheezy debianfs
+
IP=no
 +
}}
  
====Download existing====
+
{{hc|/etc/netctl/bridge|
You can download a base install tar ball. OpenVZ templates work just fine.
+
2=Description="Bridge Interface"
 +
Interface=br0
 +
Connection=bridge
 +
BindsToInterfaces=(eno1)
 +
IP=dhcp
 +
}}
  
====Using the lxc tools====
+
Make sure to first shutdown your current ethernet connections prior to bringing these up.
/usr/bin/lxc-debian {create|destroy|purge|help}
 
/usr/bin/lxc-fedora {create|destroy|purge|help}
 
  
===Creating the device nodes===
+
  # netctl stop-all
Since [[udev]] does not work within the container, you will want to make sure that a certain minimum amount of devices is created for it. This may be done with the following script:
+
  # netctl start ethernet
  #!/bin/bash
+
  # netctl start bridge
ROOT=$(pwd)
+
  # netctl enable ethernet
DEV=${ROOT}/dev
+
  # netctl enable bridge
mv ${DEV} ${DEV}.old
 
mkdir -p ${DEV}
 
  mknod -m 666 ${DEV}/null c 1 3
 
  mknod -m 666 ${DEV}/zero c 1 5
 
  mknod -m 666 ${DEV}/random c 1 8
 
  mknod -m 666 ${DEV}/urandom c 1 9
 
mkdir -m 755 ${DEV}/pts
 
mkdir -m 1777 ${DEV}/shm
 
mknod -m 666 ${DEV}/tty c 5 0
 
mknod -m 600 ${DEV}/console c 5 1
 
mknod -m 666 ${DEV}/tty0 c 4 0
 
mknod -m 666 ${DEV}/full c 1 7
 
mknod -m 600 ${DEV}/initctl p
 
mknod -m 666 ${DEV}/ptmx c 5 2
 
  
==Container configuration==
+
Now change the default config for your containers.
  
===Configuration file===
+
{{hc|/etc/lxc/default.conf|
 +
2=lxc.idmap = u 0 100000 65536
 +
lxc.idmap = g 0 100000 65536
  
The main configuration files are used to describe how to originally create a container. Though these files may be located anywhere, /etc/lxc is probably a good place.
+
lxc.net.0.type = veth
 +
lxc.net.0.link = br0
 +
lxc.net.0.flags = up
 +
}}
  
'''23/Aug/2010: Be aware that the kernel may not handle additional whitespace in the configuration file. This has been experienced on "lxc.cgroup.devices.allow" settings but may also be true on other settings. If in doubt use only one space wherever whitespace is required.'''
+
Now when you create a new container and start it, it should get an DHCP ip from the same network as the host.
  
====Basic settings====
+
=== Container creation ===
 +
Containers are built using {{ic|lxc-create}}.  With the release of lxc-3.0.0-1, upstream has deprecated locally stored templates.
  
lxc.utsname = $CONTAINER_NAME<br>
+
To build an Arch container, invoke like this:
lxc.mount = $CONTAINER_FSTAB
+
  # lxc-create -n playtime -t download -- --dist archlinux --release current --arch amd64
lxc.rootfs = $CONTAINER_ROOTFS<br>
 
lxc.network.type = veth
 
lxc.network.flags = up
 
lxc.network.link = br0
 
lxc.network.hwaddr = $CONTAINER_MACADDR
 
lxc.network.ipv4 = $CONTAINER_IPADDR
 
  lxc.network.name = $CONTAINER_DEVICENAME
 
  
=====Basic settings explained=====
+
For other distros, invoke like this and select options from the supported distros displayed in the list:
 +
# lxc-create -n playtime -t download
  
'''lxc.utsname''' : This will be the name of the cgroup for the container. Once the container is started, you should be able to see a new folder named ''/cgroup/$CONTAINER_NAME''.
+
{{Tip|Users may optionally install {{Pkg|haveged}} and [[start]] {{ic|haveged.service}} to avoid a perceived hang during the setup process while waiting for system entropy to be seeded. Without it, the generation of private/GPG keys can add a lengthy wait to the process.}}
  
Furthermore, this will also be the value returned by ''hostname'' from within the container. Assuming you have not removed access, the container may overwrite this with it's init script.
+
{{Tip|Users of [[Btrfs]] can append {{ic|-B btrfs}} to create a Btrfs subvolume for storing containerized rootfs. This comes in handy if cloning containers with the help of {{ic|lxc-clone}} command. [[ZFS]] users may use {{ic|-B zfs}}, correspondingly.}}
  
'''lxc.mount''' : This points to an fstab formatted file that is a listing of the mount points used when ''lxc-start'' is called. This file is further explained [[#Configuring fstab|further]]
+
{{Note|Users wanting the legacy templates can find them in {{AUR|lxc-templates}} or alternatively, users can build their own templates with {{AUR|distrobuilder}}.}}
  
====Terminal settings====
+
=== Container configuration ===
 +
The examples below can be used with ''privileged'' and ''unprivileged'' containers alike.  Note that for unprivileged containers, additional lines will be present by default which are not shown in the examples, including the {{ic|1=lxc.idmap = u 0 100000 65536}} and the {{ic|1=lxc.idmap = g 0 100000 65536}} values optionally defined in the [[#Enable support to run unprivileged containers (optional)]] section.
  
The following configuration is optional. You may add them to your main configuration file if you wish to login via lxc-console, or through a terminal ( e.g.: {{Keypress|Ctrl+Alt+F1}} ).
+
==== Basic config with networking ====
 +
{{Note|With the release of lxc-1:2.1.0-1, many of the configuration options have changed. Existing containers need to be updated; users are directed to the table of these changes in the [https://discuss.linuxcontainers.org/t/lxc-2-1-has-been-released/487 v2.1 release notes].}}
  
The container can be configured with virtual consoles (tty devices). These may be devices from the host that the container is given permission to use (by its configuration file) or they may be devices created locally within the container.
+
System resources to be virtualized/isolated when a process is using the container are defined in {{ic|/var/lib/lxc/CONTAINER_NAME/config}}. By default, the creation process will make a minimum setup without networking support. Below is an example config with networking:
  
The host's virtual consoles are accessed using the key sequence {{Keypress|Alt+Fn}} (or {{Keypress|Ctrl+Alt+Fn}} from within an X11 session). The left {{Keypress|Alt}} key reaches consoles 1 through 12 and the right {{Keypress|Alt}} key reaches consoles 13 through 24. Further virtual consoles may be reached by the {{Keypress|Alt+→}} key sequence which steps to the next virtual console.
+
{{hc|/var/lib/lxc/playtime/config|<nowiki>
 +
# Template used to create this container: /usr/share/lxc/templates/lxc-archlinux
 +
# Parameters passed to the template:
 +
# For additional config options, please look at lxc.container.conf(5)
  
The container's local virtual consoles may be accessed using the "lxc-console" command.
+
## default values
 +
lxc.rootfs.path = /var/lib/lxc/playtime/rootfs
 +
lxc.uts.name = playtime
 +
lxc.arch = x86_64
 +
lxc.include = /usr/share/lxc/config/common.conf
  
===== Host Virtual Consoles =====
+
## network
 +
lxc.net.0.type = veth
 +
lxc.net.0.link = br0
 +
lxc.net.0.flags = up
 +
lxc.net.0.name = eth0
 +
lxc.net.0.hwaddr = ee:ec:fa:e9:56:7d
 +
# uncomment the next two lines if static IP addresses are needed
 +
# leaving these commented will imply DHCP networking
 +
#
 +
#lxc.net.0.ipv4.address = 192.168.0.3/24
 +
#lxc.net.0.ipv4.gateway = 192.168.0.1
 +
</nowiki>}}
  
The container may access the host's virtual consoles if the host is not using them and the container's configuration allows it. Typical container configuration would deny access to all devices and then allow access to specific devices like this:
+
{{Note|The lxc.network.hwaddr entry is optional and if skipped, a random MAC address will be created automatically. It can be advantageous to define a MAC address for the container to allow the DHCP server to always assign the same IP to the container's NIC (beyond the scope of this article but worth mentioning).}}
  
  lxc.cgroup.devices.deny = a          # Deny all access to devices
+
==== Mounts within the container ====
  lxc.cgroup.devices.allow = c 4:0 rwm # /dev/tty0
+
For ''privileged'' containers, one can select directories on the host to bind mount to the container. This can be advantageous for example if the same architecture is being containerize and one wants to share pacman packages between the host and container. Another example could be shared directories. The syntax is simple:
  lxc.cgroup.devices.allow = c 4:1 rwm # /dev/tty1
 
  lxc.cgroup.devices.allow = c 4:2 rwm # /dev/tty2
 
  
For a container to be able to use a host's virtual console it must not be in use by the host. This will most likely require the host's {{ic|/etc/inittab}} to be modified to ensure no getty or other process runs on any virtual console that is to be used by the container.
+
lxc.mount.entry = /var/cache/pacman/pkg var/cache/pacman/pkg none bind 0 0
  
After editing the host's {{ic|/etc/inittab}} file, issung a {{ic|killall -HUP init}} will terminate any getty processes that are no longer configured and this will free up the virtual conosole for use by the container.
+
{{Note|This will not work without filesystem permission modifications on the host if using ''unprivileged'' containers.}}
 +
==== Xorg program considerations (optional) ====
 +
In order to run programs on the host's display, some bind mounts need to be defined so that the containerized programs can access the host's resources.  Add the following section to {{ic|/var/lib/lxc/playtime/config}}:
 +
## for xorg
 +
lxc.mount.entry = /dev/dri dev/dri none bind,optional,create=dir
 +
lxc.mount.entry = /dev/snd dev/snd none bind,optional,create=dir
 +
lxc.mount.entry = /tmp/.X11-unix tmp/.X11-unix none bind,optional,create=dir,ro
 +
lxc.mount.entry = /dev/video0 dev/video0 none bind,optional,create=file
  
Note that local virtual consoles take precedence over host virtual consoles. This is described in the next section.
+
If you still get a permission denied error in your LXC guest, then you may need to call {{ic|xhost +}} in your host to allow the guest to connect to the host's display server. Take note of the security concerns of opening up your display server by doing this.
 +
In addition you might need to add the following line
 +
lxc.mount.entry = tmpfs tmp tmpfs defaults
 +
before the bind mount lines.
  
===== Local Virtual Consoles =====
+
{{Note|This will not work if using ''unprivileged'' containers.}}
  
The number of local virtual consoles that the container has is defined in the container's configuration file (normally on the host in {{ic|/etc/lxc}}). It is defined thus:
+
==== OpenVPN considerations ====
  
  lxc.tty = n
+
Users wishing to run [[OpenVPN]] within the container are direct to either [[OpenVPN (client) in Linux containers]] and/or [[OpenVPN (server) in Linux containers]].
  
where {{ic|n}} is the number of local virtual consoles required.
+
== Managing containers ==
 +
=== Basic usage ===
 +
To list all installed LXC containers:
 +
# lxc-ls -f
  
The local virtual consoles are numbered starting at tty1 and take precedence over any of the host's virtual consoles that the container might be entitled to use. This means that, for example, if n = 2 then the container will not be able to use the host's tty1 and tty2 devices even entitled to do so by its configuration file. Setting n to 0 will prevent local virtual consoles from being created thus allowing full access to any of host's virtual consoles that the container might be entitled to use.
+
Systemd can be used to [[start]] and to [[stop]] LXCs via {{ic|lxc@CONTAINER_NAME.service}}.  [[Enable]] {{ic|lxc@CONTAINER_NAME.service}} to have it start when the host system boots.
  
===== /dev/tty Device Files =====
+
Users can also start/stop LXCs without systemd.
The container must have a tty device file (e.g. {{ic|/dev/tty1}}) for each virtual console (host or local). These can be created thus:
+
Start a container:
  # mknod -m 666 /dev/tty1 c 4 1
+
  # lxc-start -n CONTAINER_NAME
# mknod -m 666 /dev/tty2 c 4 2
 
  
and so on...
+
Stop a container:
 +
# lxc-stop -n CONTAINER_NAME
  
In the above, {{ic|c}} means character device, {{ic|4}} is the major device number (tty devices) and {{ic|1}}, {{ic|2}}, {{ic|3}}, etc., is the minor device number (specific tty device). Note that {{ic|/dev/tty0}} is special and always refers to the current virtual console.
+
To login into a container:
 +
# lxc-console -n CONTAINER_NAME
  
For further info on tty devices, read this: http://www.kernel.org/pub/linux/docs/device-list/devices.txt
+
If when login you get pts/0 and lxc/tty1 use:
 +
# lxc-console -n CONTAINER_NAME -t 0
  
'''If a virtual console's device file does not exist in the container, then the container cannot use the virtual console.'''
+
Once logged, treat the container like any other linux system, set the root password, create users, install packages, etc.
  
===== Configuring Log-In Ability =====
+
To attach to a container:
 +
# lxc-attach -n CONTAINER_NAME --clear-env
  
The container's virtual consoles may be used for login sessions if the container runs "getty" services on their tty devices. This is normally done by the container's "init" process and is configured in the container's {{ic|/etc/inittab}} file using lines like this:
+
It works nearly the same as lxc-console, but you are automatically accessing root prompt inside the container, bypassing login. Without the {{ic| --clear-env}} flag, the host will pass its own environment variables into the container (including {{ic|$PATH}}, so some commands will not work when the containers are based on another distribution).
  
  c1:2345:respawn:/sbin/agetty -8 38400 tty1 linux
+
=== Advanced usage ===
  
There is one line per device. The first part {{ic|c1}} is just a unique label, the second part defines applicable run levels, the third part tells init to start a new getty when the current one terminates and the last part gives the command line for the getty. For further information refer to {{ic|man init}}.
+
==== LXC clones ====
 +
Users with a need to run multiple containers can simplify administrative overhead (user management, system updates, etc.) by using snapshots. The strategy is to setup and keep up-to-date a single base container, then, as needed, clone (snapshot) it.  The power in this strategy is that the disk space and system overhead are truly minimized since the snapshots use an overlayfs mount to only write out to disk, only the differences in data. The base system is read-only but changes to it in the snapshots are allowed via the overlayfs.
  
If there is no getty process on a virtual console it will not be possible to log in via that virtual console. A getty is not required on a virtual console unless it is to be used to log in.
+
{{Expansion|The note needs a reference.}}
  
If a virtual console is to allow root logins it also needs to be listed in the container's {{ic|/etc/securetty}} file.
+
{{Note|overlayfs for unprivileged containers is not supported in the current mainline Arch Linux kernel due to security considerations.}}
  
===== Troubleshooting virtual consoles =====
+
For example, setup a container as outlined above.  We will call it "base" for the purposes of this guide.  Now create 2 snapshots of "base" which we will call "snap1" and "snap2" with these commands:
 +
# lxc-copy -n base -N snap1 -B overlayfs -s
 +
# lxc-copy -n base -N snap2 -B overlayfs -s
  
If lxc.tty is set to a number, n, then no host devices numbered n or below will be accessible even if the above configuration is present because they will be replaced with local virtual consoles instead.
+
{{Note|If a static IP was defined for the "base" lxc, that will need to manually changed in the config for "snap1" and for "snap2" before starting them. If the process is to be automated, a script using sed can do this automatically although this is beyond the scope of this wiki section.}}
  
A tty device file's major number will change from 4 to 136 if it is a local virtual console. This change is visible within the container but not when viewing the container's devices from the host's filesystem. This information is useful when troubleshooting.
+
The snapshots can be started/stopped like any other container. Users can optionally destroy the snapshots and all new data therein with the following command. Note that the underlying "base" lxc is untouched:
 +
# lxc-destroy -n snap1 -f
  
This can be checked from within a container thus:
+
Systemd units and wrapper scripts to manage snapshots for [[pi-hole]] and [[OpenVPN]] are available to automate the process in [https://github.com/graysky2/lxc-snapshots lxc-snapshots].
 +
 
 +
=== Converting a privileged container to an unprivileged container ===
 +
Once the system has been configured to use unprivileged containers (see, [[#Enable support to run unprivileged containers (optional)]]), {{AUR|nsexec-bzr}} contains a utility called {{ic|uidmapshift}} which is able to convert an existing ''privileged'' container to an ''unprivileged'' container to avoid a total rebuild of the image.
 +
 
 +
{{Warning|
 +
* It is recommended to backup the existing image before using this utility!
 +
* This utility will not shift UIDs and GIDs in [[ACL]], you will need to shift them on your own.
 +
}}
  
  # ls -Al /dev/tty*
+
Invoke the utility to convert over like so:
  crw------- 1 root root 136, 10 Aug 21 21:28 /dev/tty1
+
# uidmapshift -b /var/lib/lxc/foo 0 100000 65536
  crw------- 1 root root  4, 2  Aug 21 21:28 /dev/tty2
 
  
===== Pseudo Terminals =====
+
Additional options are available simply by calling {{ic|uidmapshift}} without any arguments.
  
  lxc.pseudo = 1024
+
== Running Xorg programs ==
 +
Either attach to or [[SSH]] into the target container and prefix the call to the program with the DISPLAY ID of the host's X session.  For most simple setups, the display is always 0.
  
Maximum amount of pseudo terminals that may be created in {{ic|/dev/pts}}. Currently, assuming the kernel was compiled with {{ic|CONFIG_DEVPTS_MULTIPLE_INSTANCES}}, this tells lxc-start to mount the devpts filesystem with the newinstance flag.
+
An example of running Firefox from the container in the host's display:
 +
$ DISPLAY=:0 firefox
  
====Host device access settings====
+
Alternatively, to avoid directly attaching to or connecting to the container, the following can be used on the host to automate the process:
 +
# lxc-attach -n playtime --clear-env -- sudo -u YOURUSER env DISPLAY=:0 firefox
  
lxc.cgroup.devices.deny = a # Deny all access to devices<br>
+
== Troubleshooting ==
lxc.cgroup.devices.allow = c 1:3 rwm # dev/null
 
lxc.cgroup.devices.allow = c 1:5 rwm # dev/zero<br>
 
lxc.cgroup.devices.allow = c 5:1 rwm # dev/console
 
lxc.cgroup.devices.allow = c 5:0 rwm # dev/tty
 
lxc.cgroup.devices.allow = c 4:0 rwm # dev/tty0<br>
 
lxc.cgroup.devices.allow = c 1:9 rwm # dev/urandom
 
lxc.cgroup.devices.allow = c 1:8 rwm # dev/random
 
lxc.cgroup.devices.allow = c 136:* rwm # dev/pts/*
 
lxc.cgroup.devices.allow = c 5:2 rwm # dev/pts/ptmx<br>
 
# No idea what this is .. dev/bsg/0:0:0:0 ???
 
lxc.cgroup.devices.allow = c 254:0 rwm
 
  
=====Host device access settings explained=====
+
=== Root login fails ===
  
'''lxc.cgroup.devices.deny''' : By settings this to ''a'', we are stating that the container has access to no devices unless explicitely defined within the configuration file.
+
If you get the following error when you try to login using lxc-console:
  
===Configuration file notes===
+
login: root
====At runtime /dev/ttyX devices are recreated====
+
Login incorrect
If you have enabled multiple DevPTS instances in your kernel, lxc-start will recreate ''lxc.tty'' amount of {{ic|/dev/ttyX}} devices when it is executed.
 
  
This means that you will have ''lxc.tty'' amount of pseudo ttys. If you are planning on accessing the container via a "real" terminal ({{Keypress|Ctrl+Alt+FX}}), make sure that it is a number that is inferior to ''lxc.tty''.
+
And the container's {{ic|journalctl}} shows:
  
To tell whether it has been re-created, just log in to the container via either lxc-console or SSH and perform a {{ic|ls -Al}} command on the tty. Devices with a major number of 4 are "real" tty devices whereas a major number of 136 indicates a pts.
+
pam_securetty(login:auth): access denied: tty 'pts/0' is not secure !
  
Be aware that this is only visible from within the container itself and not from the host.
+
Add {{ic|pts/0}} to the list of terminal names in {{ic|/etc/securetty}} on the '''container''' filesystem, see [http://unix.stackexchange.com/questions/41840/effect-of-entries-in-etc-securetty/41939#41939]. You can also opt to delete {{ic|/etc/securetty}} on the '''container''' to allow always root to login, see [https://github.com/systemd/systemd/issues/852].
  
====Containers have access to host's TTY nodes====
+
Alternatively, create a new user in lxc-attach and use it for logging in to the system, then switch to root.
  
If you do not properly restrict the container's access to the /dev/tty nodes, the container may have access to the host's.
+
# lxc-attach -n playtime
 +
[root@playtime]# useradd -m -Gwheel newuser
 +
[root@playtime]# passwd newuser
 +
[root@playtime]# passwd root
 +
[root@playtime]# exit
 +
# lxc-console -n playtime
 +
[newuser@playtime]$ su
  
Taking into consideration that, as previously mentioned, lxc-start recreates ''lxc.tty'' amount of /dev/tty devices, any tty nodes present in the container that are of a greater minor number than ''lxc.tty'' will be linked to the host's.
+
=== No network-connection with veth in container config===
  
=====To access the container from a host TTY=====
+
If you cannot access your LAN or WAN with a networking interface configured as '''veth''' and setup through {{ic|/etc/lxc/''containername''/config}}.
 +
If the virtual interface gets the ip assigned and should be connected to the network correctly.
 +
ip addr show veth0
 +
inet 192.168.1.111/24
 +
You may disable all the relevant static ip formulas and try setting the ip through the booted container-os like you would normaly do.
  
# On the host, verify no getty is started for that tty by checking ''/etc/inittab''.
+
Example {{ic|''container''/config}}
# In the container, start a getty for that tty.
 
  
=====To prevent access to the host TTY=====
+
...
 +
lxc.net.0.type = veth
 +
lxc.net.0.name = veth0
 +
lxc.net.0.flags = up
 +
lxc.net.0.link = {{ic|bridge}}
 +
...
  
Please have a look at the configuration statements found in [[#Host device access settings|host device access settings]].
+
And then assign your IP through your preferred method '''inside''' the container, see also [[Network configuration#Network management]].
  
Via the ''lxc.cgroup.devices.deny = a'' we are preventing access to all host level devices. And then, throuh ''lxc.cgroup.devices.allow = c 4:'''1''' rwm'' we are allowing access to the host's /dev/tty'''1'''. In the above example, simply removing all allow statements for major number 4 and minor > 1 should be sufficient.
+
=== Error: unknown command ===
  
=====To test this access=====
+
The error may happen when you type a basic command (''ls'', ''cat'', etc.) on an attached container that have different Linux distribution from the host system (e.g. Debian container in Arch Linux host system). When you attach, use the argument {{ic|--clear-env}}:
  
I may be off here, but looking at the output of the ''ls'' command below should show you both the ''major'' and ''minor'' device numbers. These are located after the user and group and represented as : 4, 2
+
# lxc-attach -n ''container_name'' --clear-env
  
# Set lxc.tty to 1
+
=== Error: Failed at step KEYRING spawning... ===
# Make there that the container has dev/tty1 and /dev/tty2
 
# ''lxc-start'' the container
 
# ''lxc-console'' into the container
 
# ''ls -Al /dev/tty''<br>crw------- 1 root root 4, 2 Dec  2 00:20 /dev/tty2
 
# ''echo "test output" > /dev/tty2''
 
# ''Ctrl+Alt+F2'' to view the host's second terminal
 
# You should see "test output" printed on the screen
 
  
====Configuration troubleshooting====
+
Services in an unprivileged container may fail with the following message
  
=====console access denied: Permission denied=====
+
some.service: Failed to change ownership of session keyring: Permission denied
 +
some.service: Failed to set up kernel keyring: Permission denied
 +
some.service: Failed at step KEYRING spawning ....: Permission denied
  
If, when executing lxc-console, you receive the error ''lxc-console: console access denied: Permission denied'' you have most likely either omitted lxc.tty or set it to 0.
+
Create a file {{ic|/etc/lxc/unpriv.seccomp}} containing
  
=====lxc-console does not provide a login prompt=====
+
{{hc|/etc/lxc/unpriv.seccomp|
 +
2
 +
blacklist
 +
[all]
 +
keyctl errno 38}}
  
Though you are reaching a tty on the container, it most likely is not running a getty. You will want to double check that you have a getty defined in the container's ''/etc/inittab'' for the specific tty.
+
Then add the following line to the container configuration '''after''' lxc.idmap
  
===Configuring fstab===
+
lxc.seccomp.profile = /etc/lxc/unpriv.seccomp
none $CONTAINER_ROOTFS/dev/pts devpts defaults 0 0
 
none $CONTAINER_ROOTFS/proc    proc  defaults 0 0
 
none $CONTAINER_ROOTFS/sys    sysfs  defaults 0 0
 
none $CONTAINER_ROOTFS/dev/shm tmpfs  defaults 0 0
 
  
This fstab is used by lxc-start when mounting the container. As such, you can define any mount that would be possible on the host such as bind mounting to the host's own filesystem. However, please be aware of any and all security implications that this may have.
+
=== Errors due to mknod changes in linux 4.18 kernel ===
  
'''Warning''' : You certainly do not want to bind mount the host's /dev to the container as this would allow it to, amongst other things, reboot the host.
+
Applications may have errors such as the following with nginx:
  
==Container Creation and Destruction==
+
nginx failes with [emerg]: open("/dev/null") failed (13: Permission denied)
  
===Creation===
+
There has been changes in linux 4.18 kernel with how mknod is handled and how it interacts with systemd's Privatedevice feature.
lxc-create -f $CONTAINER_CONFIGPATH -n $CONTAINER_NAME
 
  
''lxc-create'' will create /var/lib/lxc/$CONTAINER_NAME with a new copy of the container configuration file found in $CONTAINER_CONFIGPATH.
+
Workaround is to revert to a 4.17 version kernel until fixed upstream
  
As such, if you need to make modifications to the container's configuration file, it's advisable to modify only the original file and then perform ''lxc-destroy'' and ''lxc-create'' operations afterwards. No data will be lost by doing this.
+
* [https://github.com/lxc/lxd/issues/4950  After upgrading to 3.4 nginx won't start on container because it can't open /dev/null #4950]
  
'''Note''' : When copying the file over, lxc-create will strip all comments from the file.
+
=== Error: Failed to initialize cgroup driver / PAM unable to open pam_cgfs.so ===
  
'''Note''' : As of lxc-git from atleast ''2009-12-01'', performing lxc-create no longer splits the config file into multiple files and folders. Therefore, we only have the configuration file to worry about.
+
There is a bug in LXC 3.0.2. If you get these errors:
  
===Destruction===
+
lxc-start: $NAME: cgroups/cgfsng.c: all_controllers_found: 701 No freezer controller mountpoint found
  lxc-destroy -n $CONTAINER_NAME
+
lxc-start: $NAME: cgroups/cgroup.c: cgroup_init: 44 Failed to initialize cgroup driver                                                                         
 +
lxc-start: $NAME: start.c: lxc_init: 861 Failed to initialize cgroup driver
 +
  lxc-start: $NAME: start.c: __lxc_start: 1876 Failed to initialize container "$NAME"
 +
lxc-start: $NAME: tools/lxc_start.c: main: 330 The container failed to start
 +
lxc-start: $NAME: tools/lxc_start.c: main: 336 Additional information can be obtained by setting the --logfile and --logpriority options
  
This will delete /var/lib/lxc/$CONTAINER_NAME which only contains configuration files. No data will be lost.
+
associated with these errors in the output of journalctl:
  
==Readying the host for virtualization==
+
Aug 17 16:06:25 localhost sshd[747]: PAM unable to dlopen(/usr/lib/security/pam_cgfs.so): /usr/lib/security/pam_cgfs.so: undefined symbol: strlcat
===/etc/inittab===
+
Aug 17 16:06:25 localhost sshd[747]: PAM adding faulty module: /usr/lib/security/pam_cgfs.so
# Comment out any getty that are not required
 
  
===/etc/rc.sysinit replacement===
+
It is because of [https://github.com/lxc/lxc/issues/2556 issue 2556], which has been fixed upstream already, but has not been backported to Arch (see {{Bug|59689}}).
Since we are running in a virtual environment, a number of steps undertaken by rc.sysinit are superfluous and may even flat out fail or stall. As such, until the initscripts are made virtualization aware, this will take some hack and slash.
 
  
For now, simply replace the file :
+
[[Downgrading]] to LXC 3.0.1 may solve the problem until the fix is packaged.
#!/bin/bash
 
# Whatever is needed to clean out old daemon/service pids from your container
 
rm -f $(find /var/run -name '*pid')
 
rm -f /var/lock/subsys/*<br>
 
# Configure network settings
 
## You can either use dhcp here, manually configure your
 
## interfaces or try to get the rc.d/network script working.
 
## There have been reports that network failed in this
 
## environment.
 
ip route add default via 192.168.10.1
 
echo > /etc/resolv.conf search your-domain
 
echo >> /etc/resolv.conf nameserver 192.168.10.1<br>
 
# Initally we do not have any container originated mounts
 
rm -f /etc/mtab
 
touch /etc/mtab
 
  
===/etc/rc.conf cleanup===
+
== See also ==
You may want to remove any and all hardware related daemons from the DAEMONS line. Furthermore, depending on your situation, you may also want to remove the ''network'' daemon.
 
  
===TBC===
+
* [https://www.stgraber.org/2013/12/20/lxc-1-0-blog-post-series/ LXC 1.0 Blog Post Series]
 +
* [https://stgraber.org/2016/03/11/lxd-2-0-blog-post-series-012/ LXD 2.0: Blog post series]
 +
* [http://www.ibm.com/developerworks/linux/library/l-lxc-containers/ LXC@developerWorks]
 +
* [http://l3net.wordpress.com/tag/lxc/ LXC articles on l3net]

Latest revision as of 19:54, 12 October 2018

Linux Containers (LXC) is an operating-system-level virtualization method for running multiple isolated Linux systems (containers) on a single control host (LXC host). It does not provide a virtual machine, but rather provides a virtual environment that has its own CPU, memory, block I/O, network, etc. space and the resource control mechanism. This is provided by namespaces and cgroups features in Linux kernel on LXC host. It is similar to a chroot, but offers much more isolation.

Alternatives for using containers are systemd-nspawn, docker or rkt.

Privileged containers or unprivileged containers

LXCs can be setup to run in either privileged or unprivileged configurations.

In general, running an unprivileged container is considered safer than running a privileged container since unprivileged containers have an increased degree of isolation by virtue of their design. Key to this is the mapping of the root UID in the container to a non-root UID on the host which makes it more difficult for a hack within the container to lead to consequences on host system. In other words, if an attacker manages to escape the container, he or she should find themselves with no rights on the host.

The Arch packages currently provide out-of-the-box support for privileged containers. Unprivileged containers are only available for the system administrator with additional kernel configuration. This is due to the current Arch linux kernel shipping with user namespaces disabled for normal users. This article contains information for users to run either type of container, but additional setup is required to use unprivileged containers.

An example to illustrate unprivileged containers

To illustrate the power of UID mapping, consider the output below from a running, unprivileged container. Therein, we see the containerized processes owned by the containerized root user in the output of ps:

[root@unprivileged_container /]# ps -ef | head -n 5
UID        PID  PPID  C STIME TTY          TIME CMD
root         1     0  0 17:49 ?        00:00:00 /sbin/init
root        14     1  0 17:49 ?        00:00:00 /usr/lib/systemd/systemd-journald
dbus        25     1  0 17:49 ?        00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation
systemd+    26     1  0 17:49 ?        00:00:00 /usr/lib/systemd/systemd-networkd

On the host however, those containerized root processes are running as the mapped user (ID>100000) on the host, not as the root user on the host:

[root@host /]# lxc-info -Ssip --name sandbox
State:          RUNNING
PID:            26204
CPU use:        10.51 seconds
BlkIO use:      244.00 KiB
Memory use:     13.09 MiB
KMem use:       7.21 MiB
[root@host /]# ps -ef | grep 26204 | head -n 5
UID        PID  PPID  C STIME TTY          TIME CMD
100000   26204 26200  0 12:49 ?        00:00:00 /sbin/init
100000   26256 26204  0 12:49 ?        00:00:00 /usr/lib/systemd/systemd-journald
100081   26282 26204  0 12:49 ?        00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile --systemd-activation
100000   26284 26204  0 12:49 ?        00:00:00 /usr/lib/systemd/systemd-logind

Setup

Required software

Installing lxc and arch-install-scripts will allow the host system to run privileged lxcs.

Enable support to run unprivileged containers (optional)

Users wishing to run unprivileged containers need to complete several additional setup steps.

Firstly, a kernel is required that has support for User Namespaces (a kernel with CONFIG_USER_NS). All Arch Linux kernels have support for CONFIG_USER_NS. However, due to more general security concerns, the default Arch kernel does ship with User Namespaces enabled only for the root user. You have multiple options to create unprivileged containers:

  • Start your unprivileged containers only as root.
  • Enable the sysctl setting kernel.unprivileged_userns_clone to allow normal users to run unprivileged containers. This can be done for the current session with sysctl kernel.unprivileged_userns_clone=1 and can be made permanent with sysctl.d(5).
  • Enable the cgfs pam module by modifying /etc/pam.d/system-login to additionally contain the following line:
session optional pam_cgfs.so -c freezer,memory,name=systemd,unified

Secondly, modify /etc/lxc/default.conf to contain the following lines:

lxc.idmap = u 0 100000 65536
lxc.idmap = g 0 100000 65536

Finally, create both /etc/subuid and /etc/subgid to contain the mapping to the containerized uid/gid pairs for each user who shall be able to run the containers. The example below is simply for the root user (and systemd system unit):

/etc/subuid
root:100000:65536
/etc/subgid
root:100000:65536

Host network configuration

LXCs support different virtual network types and devices (see lxc.container.conf(5)). A bridge device on the host is required for most types of virtual networking.

LXC comes with its own NAT Bridge (lxcbr0).

Note: A NAT bridge is a standalone bridge with a private network that is not bridged to the host eth0 or a physical network. It exists as a private subnet in the host.
Tip: This is quite useful when WIFI is the only option. There have been various attempts of creating Bridges on WIFI without much success.

To use LXC's NAT Bridge you need to create its configuration file:

/etc/default/lxc-net
# Leave USE_LXC_BRIDGE as "true" if you want to use lxcbr0 for your
# containers.  Set to "false" if you'll use virbr0 or another existing
# bridge, or mavlan to your host's NIC.
USE_LXC_BRIDGE="true"

# If you change the LXC_BRIDGE to something other than lxcbr0, then
# you will also need to update your /etc/lxc/default.conf as well as the
# configuration (/var/lib/lxc/<container>/config) for any containers
# already created using the default config to reflect the new bridge
# name.
# If you have the dnsmasq daemon installed, you'll also have to update
# /etc/dnsmasq.d/lxc and restart the system wide dnsmasq daemon.
LXC_BRIDGE="lxcbr0"
LXC_ADDR="10.0.3.1"
LXC_NETMASK="255.255.255.0"
LXC_NETWORK="10.0.3.0/24"
LXC_DHCP_RANGE="10.0.3.2,10.0.3.254"
LXC_DHCP_MAX="253"
# Uncomment the next line if you'd like to use a conf-file for the lxcbr0
# dnsmasq.  For instance, you can use 'dhcp-host=mail1,10.0.3.100' to have
# container 'mail1' always get ip address 10.0.3.100.
#LXC_DHCP_CONFILE=/etc/lxc/dnsmasq.conf

# Uncomment the next line if you want lxcbr0's dnsmasq to resolve the .lxc
# domain.  You can then add "server=/lxc/10.0.3.1' (or your actual $LXC_ADDR)
# to your system dnsmasq configuration file (normally /etc/dnsmasq.conf,
# or /etc/NetworkManager/dnsmasq.d/lxc.conf on systems that use NetworkManager).
# Once these changes are made, restart the lxc-net and network-manager services.
# 'container1.lxc' will then resolve on your host.
#LXC_DOMAIN="lxc"
Tip: Make sure the bridges ip-range does not interfere with your local network.

Then we need to modify the LXC container template so our containers use our bridge:

/etc/lxc/default.conf
lxc.net.0.type = veth
lxc.net.0.link = lxcbr0
lxc.net.0.flags = up
lxc.net.0.hwaddr = 00:16:3e:xx:xx:xx

You also need to install dnsmasq which is a dependency for lxcbr0.

Enable and/or start lxc-net.service to use the bridge:

See Network bridge for more information.

Alternate Network - Bridge on same network as host

Tango-edit-cut.pngThis section is being considered for removal.Tango-edit-cut.png

Reason: Duplicates Bridge with netctl. (Discuss in Talk:Linux Containers#)

If you would like the containers to be on the same network as your host machine and not utilize NAT and firewall forwarding rules you can do the following.

You will setup a manual bridge using netctl that includes the hosts ethernet adapter. You set your ethernet to not have an ip address and instead assign an ip address to the bridge that it is binded to.

In this scenario you will not be using the lxc-net service stop it if you enabled it.

# systemctl stop lxc-net
# systemctl disable lxc-net
/etc/netctl/ethernet
Description="Manual Ethernet setup for bridge"
Interface=eno1     # your interface
Connection=ethernet
IP=no
/etc/netctl/bridge
Description="Bridge Interface"
Interface=br0
Connection=bridge
BindsToInterfaces=(eno1)
IP=dhcp

Make sure to first shutdown your current ethernet connections prior to bringing these up.

# netctl stop-all
# netctl start ethernet
# netctl start bridge
# netctl enable ethernet
# netctl enable bridge

Now change the default config for your containers.

/etc/lxc/default.conf
lxc.idmap = u 0 100000 65536
lxc.idmap = g 0 100000 65536

lxc.net.0.type = veth
lxc.net.0.link = br0
lxc.net.0.flags = up

Now when you create a new container and start it, it should get an DHCP ip from the same network as the host.

Container creation

Containers are built using lxc-create. With the release of lxc-3.0.0-1, upstream has deprecated locally stored templates.

To build an Arch container, invoke like this:

# lxc-create -n playtime -t download -- --dist archlinux --release current --arch amd64

For other distros, invoke like this and select options from the supported distros displayed in the list:

# lxc-create -n playtime -t download
Tip: Users may optionally install haveged and start haveged.service to avoid a perceived hang during the setup process while waiting for system entropy to be seeded. Without it, the generation of private/GPG keys can add a lengthy wait to the process.
Tip: Users of Btrfs can append -B btrfs to create a Btrfs subvolume for storing containerized rootfs. This comes in handy if cloning containers with the help of lxc-clone command. ZFS users may use -B zfs, correspondingly.
Note: Users wanting the legacy templates can find them in lxc-templatesAUR or alternatively, users can build their own templates with distrobuilderAUR.

Container configuration

The examples below can be used with privileged and unprivileged containers alike. Note that for unprivileged containers, additional lines will be present by default which are not shown in the examples, including the lxc.idmap = u 0 100000 65536 and the lxc.idmap = g 0 100000 65536 values optionally defined in the #Enable support to run unprivileged containers (optional) section.

Basic config with networking

Note: With the release of lxc-1:2.1.0-1, many of the configuration options have changed. Existing containers need to be updated; users are directed to the table of these changes in the v2.1 release notes.

System resources to be virtualized/isolated when a process is using the container are defined in /var/lib/lxc/CONTAINER_NAME/config. By default, the creation process will make a minimum setup without networking support. Below is an example config with networking:

/var/lib/lxc/playtime/config
# Template used to create this container: /usr/share/lxc/templates/lxc-archlinux
# Parameters passed to the template:
# For additional config options, please look at lxc.container.conf(5)

## default values
lxc.rootfs.path = /var/lib/lxc/playtime/rootfs
lxc.uts.name = playtime
lxc.arch = x86_64
lxc.include = /usr/share/lxc/config/common.conf

## network
lxc.net.0.type = veth
lxc.net.0.link = br0
lxc.net.0.flags = up
lxc.net.0.name = eth0
lxc.net.0.hwaddr = ee:ec:fa:e9:56:7d
# uncomment the next two lines if static IP addresses are needed
# leaving these commented will imply DHCP networking
#
#lxc.net.0.ipv4.address = 192.168.0.3/24
#lxc.net.0.ipv4.gateway = 192.168.0.1
Note: The lxc.network.hwaddr entry is optional and if skipped, a random MAC address will be created automatically. It can be advantageous to define a MAC address for the container to allow the DHCP server to always assign the same IP to the container's NIC (beyond the scope of this article but worth mentioning).

Mounts within the container

For privileged containers, one can select directories on the host to bind mount to the container. This can be advantageous for example if the same architecture is being containerize and one wants to share pacman packages between the host and container. Another example could be shared directories. The syntax is simple:

lxc.mount.entry = /var/cache/pacman/pkg var/cache/pacman/pkg none bind 0 0
Note: This will not work without filesystem permission modifications on the host if using unprivileged containers.

Xorg program considerations (optional)

In order to run programs on the host's display, some bind mounts need to be defined so that the containerized programs can access the host's resources. Add the following section to /var/lib/lxc/playtime/config:

## for xorg
lxc.mount.entry = /dev/dri dev/dri none bind,optional,create=dir
lxc.mount.entry = /dev/snd dev/snd none bind,optional,create=dir
lxc.mount.entry = /tmp/.X11-unix tmp/.X11-unix none bind,optional,create=dir,ro
lxc.mount.entry = /dev/video0 dev/video0 none bind,optional,create=file

If you still get a permission denied error in your LXC guest, then you may need to call xhost + in your host to allow the guest to connect to the host's display server. Take note of the security concerns of opening up your display server by doing this. In addition you might need to add the following line

lxc.mount.entry = tmpfs tmp tmpfs defaults

before the bind mount lines.

Note: This will not work if using unprivileged containers.

OpenVPN considerations

Users wishing to run OpenVPN within the container are direct to either OpenVPN (client) in Linux containers and/or OpenVPN (server) in Linux containers.

Managing containers

Basic usage

To list all installed LXC containers:

# lxc-ls -f

Systemd can be used to start and to stop LXCs via lxc@CONTAINER_NAME.service. Enable lxc@CONTAINER_NAME.service to have it start when the host system boots.

Users can also start/stop LXCs without systemd. Start a container:

# lxc-start -n CONTAINER_NAME

Stop a container:

# lxc-stop -n CONTAINER_NAME

To login into a container:

# lxc-console -n CONTAINER_NAME

If when login you get pts/0 and lxc/tty1 use:

# lxc-console -n CONTAINER_NAME -t 0

Once logged, treat the container like any other linux system, set the root password, create users, install packages, etc.

To attach to a container:

# lxc-attach -n CONTAINER_NAME --clear-env

It works nearly the same as lxc-console, but you are automatically accessing root prompt inside the container, bypassing login. Without the --clear-env flag, the host will pass its own environment variables into the container (including $PATH, so some commands will not work when the containers are based on another distribution).

Advanced usage

LXC clones

Users with a need to run multiple containers can simplify administrative overhead (user management, system updates, etc.) by using snapshots. The strategy is to setup and keep up-to-date a single base container, then, as needed, clone (snapshot) it. The power in this strategy is that the disk space and system overhead are truly minimized since the snapshots use an overlayfs mount to only write out to disk, only the differences in data. The base system is read-only but changes to it in the snapshots are allowed via the overlayfs.

Tango-view-fullscreen.pngThis article or section needs expansion.Tango-view-fullscreen.png

Reason: The note needs a reference. (Discuss in Talk:Linux Containers#)
Note: overlayfs for unprivileged containers is not supported in the current mainline Arch Linux kernel due to security considerations.

For example, setup a container as outlined above. We will call it "base" for the purposes of this guide. Now create 2 snapshots of "base" which we will call "snap1" and "snap2" with these commands:

# lxc-copy -n base -N snap1 -B overlayfs -s
# lxc-copy -n base -N snap2 -B overlayfs -s
Note: If a static IP was defined for the "base" lxc, that will need to manually changed in the config for "snap1" and for "snap2" before starting them. If the process is to be automated, a script using sed can do this automatically although this is beyond the scope of this wiki section.

The snapshots can be started/stopped like any other container. Users can optionally destroy the snapshots and all new data therein with the following command. Note that the underlying "base" lxc is untouched:

# lxc-destroy -n snap1 -f

Systemd units and wrapper scripts to manage snapshots for pi-hole and OpenVPN are available to automate the process in lxc-snapshots.

Converting a privileged container to an unprivileged container

Once the system has been configured to use unprivileged containers (see, #Enable support to run unprivileged containers (optional)), nsexec-bzrAUR contains a utility called uidmapshift which is able to convert an existing privileged container to an unprivileged container to avoid a total rebuild of the image.

Warning:
  • It is recommended to backup the existing image before using this utility!
  • This utility will not shift UIDs and GIDs in ACL, you will need to shift them on your own.

Invoke the utility to convert over like so:

# uidmapshift -b /var/lib/lxc/foo 0 100000 65536

Additional options are available simply by calling uidmapshift without any arguments.

Running Xorg programs

Either attach to or SSH into the target container and prefix the call to the program with the DISPLAY ID of the host's X session. For most simple setups, the display is always 0.

An example of running Firefox from the container in the host's display:

$ DISPLAY=:0 firefox

Alternatively, to avoid directly attaching to or connecting to the container, the following can be used on the host to automate the process:

# lxc-attach -n playtime --clear-env -- sudo -u YOURUSER env DISPLAY=:0 firefox

Troubleshooting

Root login fails

If you get the following error when you try to login using lxc-console:

login: root
Login incorrect

And the container's journalctl shows:

pam_securetty(login:auth): access denied: tty 'pts/0' is not secure !

Add pts/0 to the list of terminal names in /etc/securetty on the container filesystem, see [1]. You can also opt to delete /etc/securetty on the container to allow always root to login, see [2].

Alternatively, create a new user in lxc-attach and use it for logging in to the system, then switch to root.

# lxc-attach -n playtime
[root@playtime]# useradd -m -Gwheel newuser
[root@playtime]# passwd newuser
[root@playtime]# passwd root
[root@playtime]# exit
# lxc-console -n playtime
[newuser@playtime]$ su

No network-connection with veth in container config

If you cannot access your LAN or WAN with a networking interface configured as veth and setup through /etc/lxc/containername/config. If the virtual interface gets the ip assigned and should be connected to the network correctly.

ip addr show veth0 
inet 192.168.1.111/24

You may disable all the relevant static ip formulas and try setting the ip through the booted container-os like you would normaly do.

Example container/config

...
lxc.net.0.type = veth
lxc.net.0.name = veth0
lxc.net.0.flags = up
lxc.net.0.link = bridge
...

And then assign your IP through your preferred method inside the container, see also Network configuration#Network management.

Error: unknown command

The error may happen when you type a basic command (ls, cat, etc.) on an attached container that have different Linux distribution from the host system (e.g. Debian container in Arch Linux host system). When you attach, use the argument --clear-env:

# lxc-attach -n container_name --clear-env

Error: Failed at step KEYRING spawning...

Services in an unprivileged container may fail with the following message

some.service: Failed to change ownership of session keyring: Permission denied
some.service: Failed to set up kernel keyring: Permission denied
some.service: Failed at step KEYRING spawning ....: Permission denied

Create a file /etc/lxc/unpriv.seccomp containing

/etc/lxc/unpriv.seccomp
2
blacklist
[all]
keyctl errno 38

Then add the following line to the container configuration after lxc.idmap

lxc.seccomp.profile = /etc/lxc/unpriv.seccomp

Errors due to mknod changes in linux 4.18 kernel

Applications may have errors such as the following with nginx:

nginx failes with [emerg]: open("/dev/null") failed (13: Permission denied)

There has been changes in linux 4.18 kernel with how mknod is handled and how it interacts with systemd's Privatedevice feature.

Workaround is to revert to a 4.17 version kernel until fixed upstream

Error: Failed to initialize cgroup driver / PAM unable to open pam_cgfs.so

There is a bug in LXC 3.0.2. If you get these errors:

lxc-start: $NAME: cgroups/cgfsng.c: all_controllers_found: 701 No freezer controller mountpoint found
lxc-start: $NAME: cgroups/cgroup.c: cgroup_init: 44 Failed to initialize cgroup driver                                                                          
lxc-start: $NAME: start.c: lxc_init: 861 Failed to initialize cgroup driver
lxc-start: $NAME: start.c: __lxc_start: 1876 Failed to initialize container "$NAME"
lxc-start: $NAME: tools/lxc_start.c: main: 330 The container failed to start
lxc-start: $NAME: tools/lxc_start.c: main: 336 Additional information can be obtained by setting the --logfile and --logpriority options

associated with these errors in the output of journalctl:

Aug 17 16:06:25 localhost sshd[747]: PAM unable to dlopen(/usr/lib/security/pam_cgfs.so): /usr/lib/security/pam_cgfs.so: undefined symbol: strlcat
Aug 17 16:06:25 localhost sshd[747]: PAM adding faulty module: /usr/lib/security/pam_cgfs.so

It is because of issue 2556, which has been fixed upstream already, but has not been backported to Arch (see FS#59689).

Downgrading to LXC 3.0.1 may solve the problem until the fix is packaged.

See also