Search domain support Signed-off-by: Erik Hollensbe <firstname.lastname@example.org>
|1 year ago|
|cmd||1 year ago|
|db||1 year ago|
|dhcpd||1 year ago|
|proto||1 year ago|
|testutil||1 year ago|
|version||1 year ago|
|.gitignore||1 year ago|
|.golangci.yml||1 year ago|
|LICENSE||1 year ago|
|Makefile||1 year ago|
|README.md||1 year ago|
|VERSION||1 year ago|
|box.rb||1 year ago|
|entrypoint.sh||1 year ago|
|example.conf||1 year ago|
|go.mod||1 year ago|
|go.sum||1 year ago|
This is a DHCP service/daemon with very few features. It provides basic dynamic IPv4 pool allocation as well as persistent, static leases. iPXE support does not exist yet, but is planned.
One thing Light DHCPd offers that is novel, is a remote control plane powered over GRPC, authenticated and encrypted by TLS client certificates. This control plane can be embedded into your orchestration code or you can use the provided command-line tool to manipulate it from your shell. There is a golang client, and the protobufs are included in the source tree if you wish to generate clients for other languages.
There is a small configuration file for managing dynamic leases and the overall network parameters (resolver, gateway).
Light DHCPd has been powering my network for over a week!
(Seriously, you do not want to use this in production yet.)
make shell: This runs a docker shell. You can do a few things in here:
make test: This is context-dependent and will run properly in the container or outside of it, running the unit and functional tests in a container.
make interfaces: This sets up some dummy interfaces and plumbs them through a bridge; afterwards
veth1will be available for running a
veth3will be available for running a
make start: will start the dhcpd.
make stop: stops it.
make get-ip: issues a ISC
dhclientlaunch against the second veth pair to get an IP, allowing you to test interaction with a real client.
NOTE: The following instructions install https://github.com/box-builder/box
on first run to build the images, will require
sudo for that (but nothing
else). If you don't want to run
sudo to install box, install it in your
$PATH somewhere and try again.
Assuming you're not crazy enough to try this on your own network, try this at your shell instead:
$ make shell # <inside of container> $ make interfaces $ make install $ sudo ldhcpd veth1 example.conf & $ sudo dhclient -1 -v -d veth3 # ^C it to stop it $ ip addr # veth3 -> 10.0.20.50
If you want to boot the control plane only, without serving DHCP, try the
Config File Rundown
# # DNS servers # dns_servers: - 10.0.0.1 - 220.127.116.11 # # network gateway # gateway: 10.0.20.1 # # Search domains (if not specified, the client will not get a default) # search_domains: - internal # # Dynamic Range of IPs to use in dynamic lease hand-outs, IP inclusive. # dynamic_range: from: 10.0.20.50 to: 10.0.20.100 # # Lease parameters: # # The duration is the duration of the lease; no other allocation can affect the # IP you will get back while this lease is obtained. # # The grace period is the maximum amount of time the IP is available to the mac # address; it is added to the duration. If another mac comes in and there are # no available IPs, addresses in the grace period may be reclaimed to make # room. # lease: duration: 24h grace_period: 8h
Making your certificate authority
We use mkcert to generate our certs.
You can use this script to generate a very basic CA that operates over
localhost. Depending on your circumstances, you may need to run it as
This code also installs it into the paths expected by default in the daemon and client; be mindful of permissions and directory names!
CAROOT=/etc/ldhcpd mkcert -install # CA file will be /etc/ldhcpd/rootCA.pem CAROOT=/etc/ldhcpd mkcert -cert-file /etc/ldhcpd/server.pem -key-file /etc/ldhcpd/server.key localhost 127.0.0.1 CAROOT=/etc/ldhcpd mkcert -client -cert-file /etc/ldhcpd/client.pem -key-file /etc/ldhcpd/client.key localhost 127.0.0.1
ldhcpd will suss out your subnet block from the interface you tell it to listen
on. It is very important your interface is configured correctly. We make
some checks, but no guarantees! Most importantly, ensure the subnet for ip is
configured properly. Subnetting is a bit beyond this document's scope, but
this is typically done by associating the IP with a CIDR address space that is
2: ens18: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000 link/ether a6:80:e0:5e:c5:46 brd ff:ff:ff:ff:ff:ff inet 10.0.0.2/24 brd 10.0.0.255 scope global ens18 valid_lft forever preferred_lft forever inet6 2001:1::1/64 scope global valid_lft forever preferred_lft forever inet6 fe80::a480:e0ff:fe5e:c546/64 scope link valid_lft forever preferred_lft forever
10.0.0.2/24 will get selected here to serve. Only one address like this
may be configured for now. We may eventually address this, but for now if you
need to serve two subnets, start two of
ldhcpd on different interfaces.
These are the items planned for the near future of this project:
- Per-Lease DNS and Gateway parameters
- Hostname support:
- Pushing hostnames
- Recording hostnames from clients
- Better, easier to use bridge for the GRPC client
- PXE booting support
- maybe with TFTP baked-in?
- github.com/insomniacslk/dhcp and https://github.com/krolaw/dhcp4 for the dhcp4 protocol work, thanks to the authors, This tool would be much less useful without it.
- https://github.com/jinzhu/gorm and https://github.com/mattn/go-sqlite3 for the database work.
- https://google.golang.org/grpc for the control plane protocol.
- https://github.com/box-builder/box which is a mruby-based docker image builder.
Erik Hollensbe email@example.com