Using static IPs with LXD

LXD
21

Hi Simos, thank you, interesting… I’ll have a play around… I want to ssh into various containers on the same host, so I suspect I’ll need to have different ports for ssh on each container…

22

You either use different ports for each container.

Or, you can set up a bastion host for SSH. You can google for the bastion host and get some nice tutorials.

1 Like
23

Great, thank you…

25

The default expiration time for the “managed” lxdbr0 is actually 1h, check the output of:

lxc network get lxdbr0 ipv4.dhcp.expiry

If you check the parameters with which a dnsmasq instance is started you’ll notice the lease expiration set as part of the “--dhcp-range” parameter. According to dnsmasq man pages the parameter accepts a value of “infinite” so this should work for an infinite lease expiration time:

lxc network set lxdbr0 ipv4.dhcp.expiry=infinite

I presume that @stgraber might come up with some edge cases where using this is not advisable or cases when LXD might reinitialize /var/snap/lxd/common/lxd/networks/lxdbr0/dnsmasq.leases leading to loss of “infinite” leases.

26

That worked for me perfectly to set the IP address to a fixed value for the container, but since I’ve updated to LXD 5.6 / 5.7 that commands are no longer working:

lxc network attach lxdbr0 wireguard eth0 eth0
Error: Failed add validation for device “eth0”: Instance DNS name “wireguard” already used on network

lxc config set wireguard eth0 ipv4.address=10.0.5.102
Error: Invalid config: Unknown configuration key: eth0

Do you have any ideas?

27

You should be using lxc config device set ...
This hasn’t changed in LXD 5.7.

28

Thanks! This was the solution for me as well. :wink:

29

In case you need to create an instance from scratch (the way I understand it):

# determine the bridge's ip
$ lxc network get lxdbr0 ipv4.address
10.99.10.1/24
# as such the network's address range is apparently 10.99.10.0/24

# create an instance
$ lxc init images:debian/12 d12

# attach lxdbr0 to d12 as a NIC device
$ lxc network attach lxdbr0 d12 eth0 eth0

# set ip for the device
$ lxc config device set d12 eth0 ipv4.address 10.99.10.2

$ lxc start d12
$ lxc list
+------+---------+-------------------+------+-----------+-----------+
| NAME |  STATE  |       IPV4        | IPV6 |   TYPE    | SNAPSHOTS |
+------+---------+-------------------+------+-----------+-----------+
| d12  | RUNNING | 10.99.10.2 (eth0) | ...  | CONTAINER | 0         |
+------+---------+-------------------+------+-----------+-----------+
$ lxc delete --force d12

But lxc network attach is a shortcut for lxc config device add, as such:

$ lxc init images:debian/12 d12
$ lxc config device add d12 eth0 nic network=lxdbr0 ipv4.address=10.99.10.2
$ lxc start d12
$ lxc list
+------+---------+-------------------+------+-----------+-----------+
| NAME |  STATE  |       IPV4        | IPV6 |   TYPE    | SNAPSHOTS |
+------+---------+-------------------+------+-----------+-----------+
| d12  | RUNNING | 10.99.10.2 (eth0) | ...  | CONTAINER | 0         |
+------+---------+-------------------+------+-----------+-----------+
$ lxc delete --force d12

Or even this way:

$ lxc launch -d eth0,ipv4.address=10.99.10.2 images:debian/12 d12
$ lxc list
+------+---------+-------------------+------+-----------+-----------+
| NAME |  STATE  |       IPV4        | IPV6 |   TYPE    | SNAPSHOTS |
+------+---------+-------------------+------+-----------+-----------+
| d12  | RUNNING | 10.99.10.2 (eth0) | ...  | CONTAINER | 0         |
+------+---------+-------------------+------+-----------+-----------+
$ lxc delete --force d12

I didn’t find a way to determine the network’s ip range, but determining it via the bridge’s ip is probably fine. I’m not sure what’s the difference between a device name and a interface name (lxc network attach). And generally I’m not an expert at LXD, so I might be missing something.

Useful links:

UPD Particularly, I don’t understand why exactly I can’t use eth1 in place of eth0. E.g. how do I create an instance without eth0 and with eth1 that has a static ip? Or with eth0 (dynamic ip) and eth1 (static ip)? Not that that is important. Rather out of curiosity. To better understand how it works.

30

I got static IPs working with routed like this:

# lxc profile show default
config:
  cloud-init.network-config: |
    version: 2
    ethernets:
      eth0:
        nameservers:
          addresses: [8.8.8.8]
description: Default LXD profile
devices:
  eth0:
    nictype: routed
    parent: eno1np0
    type: nic
name: default

For each container:

lxc init images:debian/12/cloud c0
lxc config device override c0 eth0 ipv4.address=1.2.3.4
lxc start c0
lxc exec c0 -- systemctl stop systemd-networkd
lxc exec c0 -- systemctl disable systemd-networkd
lxc exec c0 -- sed -i 's/#DNS=/DNS=8.8.8.8/g' /etc/systemd/resolved.conf
lxc restart c0
1 Like