[Bug 1770082] Re: systemd-networkd not renaming devices on boot

[Mathieu Trudel-Lapierre]

There are a couple of pieces in play here.

One aspect is that we don't really want to write the .link files for
systemd-networkd to /lib or get anything from /run into the initrd --
that defeats the purpose of netplan's config being dynamic.

The second aspect is that depending on how the systems are deployed
(MAAS? cloud images? openstack? VM? hardware?) changes the behavior a
bit. Some USB-based network interfaces are simply "unaffected" and
rename just fine at boot. Other systems may be renaming the interfaces,
but seeing cloud-init re-change the name *back* to the previous value
(this has been observed for MAAS-deployed systems). This is why I added
cloud-init to affected packages -- cloud-init should not be second-
guessing the network layer and attempting to do renames / to run
udevadm, except *maybe* at first boot/when the instance data is created.
I believe it should just leave it alone completely, and let
netplan/systemd/NM deal with the interfaces themselves (if anything
needs to be poked, it's arguably a bug in the backend).

Another issue is that systemd is also second-guessing configuration. If
we set things to be renamed, we do want them to be renamed. Now, this is
a bit more up to discussion upstream, but I do think systemd should
always rename if configuration tells it to (via .link files or udev
rules -- .link files are "cleaner", easier to write). Users renaming
their interfaces manually via 'ip link dev X name Y' are doing so on
their own, and the change would not be persistent across a reboot --
renaming in configuration is expected to work normally.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to systemd in Ubuntu.
https://bugs.launchpad.net/bugs/1770082

Title:
  systemd-networkd not renaming devices on boot

Status in netplan:
  Confirmed
Status in cloud-init package in Ubuntu:
  Confirmed
Status in systemd package in Ubuntu:
  Confirmed

Bug description:
  === systemd issue ===

  Renaming devices doesn't seem to work.

  If I disable all other network configuration and create
  /etc/systemd/network/10-network.link with:

  [Match]
  MACAddress=52:54:00:c1:c9:bb

  [Link]
  Name=myiface3

  I expect this to cause the device with that MAC address to be renamed
  to  myiface3. However, when I reboot, I instead see:

  $ ip l
  1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
      link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
  2: ens3: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
      link/ether 52:54:00:c1:c9:bb brd ff:ff:ff:ff:ff:ff

  The device is not renamed.

  This link file is pretty much identical to Example 2 in
  https://www.freedesktop.org/software/systemd/man/systemd.link.html.

  The renaming does work if I boot with net.ifnames=0, and oddly, it
  also works if I unbind the device and rebind it as netplan apply does.
  No setting of NamePolicy seems to help.

  === Original Bug ==

  'set-name:' doesn't change the name of a network interface on boot, it
  only works when you do netplan apply.

  Say I take this 50-cloud-init.yaml file:

  # This file is generated from information provided by
  # the datasource.  Changes to it will not persist across an instance.
  # To disable cloud-init's network configuration capabilities, write a file
  # /etc/cloud/cloud.cfg.d/99-disable-network-config.cfg with the following:
  # network: {config: disabled}
  network:
      version: 2
      ethernets:
          ens3:
              dhcp4: true
              match:
                  macaddress: 52:54:00:de:bd:f6
              set-name: ens3

  Say I change set-name to 'myiface3' and reboot. I expect that the
  device will be called myiface3 and brought up fine with dhcp. However,
  instead I see:

  $ ip a
  1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
      link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
      inet 127.0.0.1/8 scope host lo
         valid_lft forever preferred_lft forever
      inet6 ::1/128 scope host
         valid_lft forever preferred_lft forever
  2: ens3: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
      link/ether 52:54:00:de:bd:f6 brd ff:ff:ff:ff:ff:ff

  The name has not been changed, and the device has not been brought up.

  If I run netplan apply however, I see the following:

  1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
      link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
      inet 127.0.0.1/8 scope host lo
         valid_lft forever preferred_lft forever
      inet6 ::1/128 scope host
         valid_lft forever preferred_lft forever
  3: myiface3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
      link/ether 52:54:00:de:bd:f6 brd ff:ff:ff:ff:ff:ff
      inet 192.168.122.151/24 brd 192.168.122.255 scope global dynamic myiface3
         valid_lft 3575sec preferred_lft 3575sec
      inet6 fe80::5054:ff:fede:bdf6/64 scope link
         valid_lft forever preferred_lft forever

  So names are successfully changed with netplan apply.

  This seems to be some udev-related timing or priority issue that I'm
  still trying to hunt down.

  This breaks some forms of migration in certain cloud environments.

To manage notifications about this bug go to:
https://bugs.launchpad.net/netplan/+bug/1770082/+subscriptions