LXE is a shim of the Kubernetes Container Runtime Interface for LXD. This project is currently under heavy development, expect incompatible changes.
You need to have LXD installed, which packages are officially only available via snap. Debian is working on a LXD deb package, other distros might be as well. A LXD built by source is also supported.
Head over to the releases page for available builds.
Please follow these steps carefully. Some parameters and arguments depend on how you installed LXD.
Make sure that you have LXD running and the LXD-client's remote configuration file exists (e.g. by running lxc list
once), LXE will need that later.
LXE can be run as a non-privileged user, so give it access to lxd's socket. When using the network-plugin cni root permissions are required.
The most important LXE options are the following:
--lxd-remote-config string Path to the LXD remote config. (guessed by default)
--lxd-socket string Path of the socket where LXD provides it's API. (guessed by default)
--network-plugin string The network plugin to use. 'bridge' manages the lxd bridge defined in --bridge-name. 'cni' uses container network interface to attach interfaces using a configuration defined in --cni-conf-dir. (default "bridge")
--socket string Path of the socket where it should provide the runtime and image service to kubelet. (default "/run/lxe.sock")
We recommend to use CNI as the network plugin as it offers more flexibility and integration to common kubernetes network setups. But for sure you can use the currently default network plugin, which uses lxd's integrated networking, and build kubernetes cluster networking around it.
The CNI plugin is selected by passing the --network-plugin=cni
option. The CNI configuration is read from within --cni-conf-dir
(default /etc/cni/net.d) and uses that file to set up each pod’s network. The CNI configuration file must match the CNI specification, and any required CNI plugins referenced by the configuration must be present in --cni-bin-dir
(default /opt/cni/bin).
If there are multiple CNI configuration files in the directory, the first configuration file by name in lexicographic order is used. Keep in mind you can also chain several plugins using a conflist file. Example configuration /etc/cni/net.d/10-mynet.conf
:
{
"cniVersion": "0.3.1",
"name": "mynet",
"type": "bridge",
"bridge": "cni0",
"isGateway": true,
"ipMasq": true,
"ipam": {
"type": "host-local",
"ranges": [
[
{
"subnet": "10.22.0.0/16",
"rangeStart": "10.22.0.50",
"rangeEnd": "10.22.0.100",
"gateway": "10.22.0.1"
}
]
],
"routes": [
{
"dst": "0.0.0.0/0"
}
]
}
}
For all options, consider looking into lxe --help
.
You might want to use --log-level info
for some feedback, otherwise the daemon is pretty silent when no warnings or errors occur:
lxe --network-plugin cni --log-level info
You should be greeted with:
WARNING[07-20|17:11:18.042] starting lxe... packagename=lxe version=0.0.0 gitcommit=... gittreestate=dirty buildnumber=undef builddate="..."
INFO [07-20|17:11:18.051] Connected to LXD lxd-socket=/var/lib/lxd/unix.socket
INFO [07-20|17:11:18.068] Started lxe CRI shim socket=/run/lxe.sock
INFO [07-20|17:11:18.068] Started streaming server endpoint=":44124" baseurl="http://10.249.100.169:44124"
Aside from command-line parameters, LXE also supports configuration files in various file formats (YAML, JSON, TOML, etc.) and environment variables. All parameters are automatically detected depending on rules for each configuaration type:
--lxd-socket
is LXD_SOCKET
--lxd-socket
is {"lxd": {"socket": ...}}
You can also print the currently loaded configuration with lxd config show yaml
(or any other supported extension). This only shows what configuration would be loaded if you would run lxe, it does not retrieve the current running configuration in case you have already an lxe process running!
$ lxe config show toml
config = ""
socket = "/run/lxe.sock"
[bridge]
name = "lxdbr0"
[bridge.dhcp]
range = ""
...
If no config argument is provided via command-line or environment, it will automatically look for a file lxe.<ext>
in ~/.local/lxe/
and /etc/lxe/
(The folder lxe
can be customized with the make variable PACKAGE_NAME
)
You can also combine all these variants. Command-line parameters have precedence over environment variables, which have precedence over configuration file settings, those in turn have precedence over defaults. Please be aware you can't set the config variable in a config file, it has no effect. Once a variable is set in any way, even if empty, the default is overridden.
Now that you have LXE running on your system you can define the LXE socket as CRI endpoint in kubelet. You'll have to define the following options --container-runtime=remote
and --container-runtime-endpoint=unix:///run/lxe.sock
and your kubelet should be able to connect to your LXE socket.
LXE requires golang 1.18.
Simply run:
go install github.com/automaticserver/lxe/cmd/lxe@latest
And the binary will be located in $GOPATH/bin
Clone this repo to your wished location. Build this project using the following command, which will give you the binary in ./bin/
make build
You can also run the program directly using:
make run lxe -- --log-level info --network-plugin cni
There are also tests and lints available.
make test
make lint
For testing with critest
from cri-tools have a look in the testing documentation.
To list all available make targets have a look at help target make help
.
Please have a look at the multi node cluster guide
Bug reports can be filed at the github issue tracker
A lot of options are missing and not yet implemented from the Kubernetes PodSpec. Limitations and decisions of the current state are described in the development preview FAQ.