| Line 166: | Line 166: | ||
</pre> | </pre> | ||
You'll find two physical network cards (<code>enp1s0f0</code> and <code>enp1s0f1</code>) each attached to a bridge (<code>br0</code> and <code>br1</code>, respectively). The physical network adapters show up as 'degraded' because they don't have network addresses directly assigned -- that assignment is done on the bridge. | You'll find two physical network cards (<code>enp1s0f0</code> and <code>enp1s0f1</code>) each attached to a bridge (<code>br0</code> and <code>br1</code>, respectively). The physical network adapters show up as 'degraded' because they don't have network addresses directly assigned -- that assignment is done on the bridge. The <code>gre0</code> and <code>gretap0</code> devices are created automatically to handle the gre tunnel <code>gre-colocation</code>. There are also two VLANs configured within systemd and attached to a bridge. The <code>tun</code> interfaces are OpenVPN interfaces and they are not configured by systemd-networkd (hence the `unmanaged` setup). | ||
===Further Reading=== | ===Further Reading=== | ||
Revision as of 02:37, 29 September 2015
This is a placeholder page for the discussion of what the Cloud image (Base at first, Atomic to follow) requires from a networking stack on it's images.
- configure DHCP, renew dhcp leases
- configure with cloud-init
- configure with traditional RH-ecosystem ifcfg-eth0 file (log warnings for unsupported options)
Questions!
- Q: any need to support more than one interface? - Q: if single interface, should we standardize on "eth0"" - Q: dns configuration? - Q: support static IP (via cloud-init?)
Several use cases in EC2 involve instances with multiple network interfaces and occasionally multiple addresses on a single interface. I do not believe all of those need to be configured automatically, but it ought to remain possible for a user/script/etc to make network configuration changes that persist across reboots.
.png){kind=small}
mhayden is doing some rambling on systemd-networkd here since maxamillion asked so nicely.
systemd-networkd use cases
Here are some sample use cases for systemd-networkd and example configurations.
Simple DHCP on a single interface
For an interface eth0, a single .network file is needed:
# cat /etc/systemd/network/eth0.network [Match] Name=eth0 [Network] DHCP=yes
Static address on a single interface
For an interface eth0, a single .network file is needed:
# cat /etc/systemd/network/eth0.network [Match] Name=eth0 [Network] Address=192.168.0.50/24 Address=2001:db8:dead:beef::/64 # These are optional but worth mentioning DNS=8.8.8.8 DNS=8.8.4.4 NTP=pool.ntp.org
You can also split up the addresses into separate blocks:
# cat /etc/systemd/network/eth0.network [Match] Name=eth0 [Network] DNS=8.8.8.8 DNS=8.8.4.4 NTP=pool.ntp.org [Address] Address=192.168.0.50/24 [Address] Address=2001:db8:dead:beef::/64
Or add static routes:
# cat /etc/systemd/network/eth0.network [Match] Name=eth0 [Network] DNS=8.8.8.8 DNS=8.8.4.4 NTP=pool.ntp.org [Address] Address=192.168.0.50/24 [Address] Address=2001:db8:dead:beef::/64 [Route] Destination=10.0.10.0/24 Gateway=192.168.50.1 [Route] Destination=10.0.20.0/24 Gateway=192.168.50.1
Do DHCP on all network devices
You can use wildcards almost anywhere in the [Match] block. For example, this will cause systemd-networkd to do DHCP on all interfaces:
[Match] Name=eth* [Network] DHCP=yes
Bridging
Let's consider an example where we have eth0 and we want to add it to a bridge. This could be handy for servers where you want to build containers or virtual machines and attach them to the network bridge.
Start by setting up our bridge interface, br0:
# cat /etc/systemd/network/br0.netdev [NetDev] Name=br0 Kind=bridge
Now that we have a bridge device, let's configure the network for the bridge:
# cat /etc/systemd/network/br0.network [Match] Name=br0 [Network] IPForward=yes DHCP=yes
The IPForward=yes will take care of the sysctl forwarding setting for us (net.ipv4.conf.br0.forwarding = 1) automatically when the interface comes up.
Now, let's take the ethernet adapter and add it to the bridge:
# cat /etc/systemd/network/eth0.network [Match] Name=eth0 [Network] Bridge=br0
Simply reboot the system and it will come up with eth0 as a port on br0.
Status & Diagnostics
All of the output from systemd-networkd will appear in your system journal. Any errors when setting up interfaces or configuring routes will be printed there. The networkctl command allows you to check your network devices at a glance. Here's an example of a fairly complicated network setup:
# networkctl
IDX LINK TYPE OPERATIONAL SETUP
1 lo loopback carrier unmanaged
2 enp3s0 ether off unmanaged
3 enp1s0f0 ether degraded configured
4 enp1s0f1 ether degraded configured
5 br1 ether routable configured
6 br0 ether routable configured
7 gre0 ipgre off unmanaged
8 gretap0 ether off unmanaged
9 gre-colocation ipgre routable configured
12 vlan100 ether routable configured
13 tun1 none routable unmanaged
14 tun0 none routable unmanaged
15 vlan200 ether routable configured
You'll find two physical network cards (enp1s0f0 and enp1s0f1) each attached to a bridge (br0 and br1, respectively). The physical network adapters show up as 'degraded' because they don't have network addresses directly assigned -- that assignment is done on the bridge. The gre0 and gretap0 devices are created automatically to handle the gre tunnel gre-colocation. There are also two VLANs configured within systemd and attached to a bridge. The tun interfaces are OpenVPN interfaces and they are not configured by systemd-networkd (hence the unmanaged setup).