Disabling IPv6 causes network interface to stay in configuring status

[basvdlei]

Issue Report

Bug

Since our beta machines have been upgraded to version 1068.2.0, they alway boot up in a degraded status because unit systemd-networkd-wait-online fails.

networkctl shows our (statically configured) interface as configuring and never gets the configured status. But the configuration has actually been applied and networking does work.

Running the systemd-networkd-wait-online command manually, it never returns.

We have our servers configured with static networking:

write_files:
- path: /etc/systemd/network/static.network
  permissions: "0644"
  content: |
    [Match]
    Name=en*

    [Network]
    Address=xxx.xx.xx.xxx/xx
    Gateway=xxx.xx.xx.xxx
    DNS=xxx.xx.xx.xxx
    DNS=xxx.xx.xx.xxx

And have IPv6 disabled with a sysctl drop-in:

write_files:
- path: /etc/sysctl.d/10-disable-ipv6.conf
  permissions: "0644"
  owner: root
  content: |
    net.ipv6.conf.all.disable_ipv6 = 1

When I re-enable IPv6 by removing the drop-in, the system boots up normally.

CoreOS Version

NAME=CoreOS
ID=coreos
VERSION=1068.2.0
VERSION_ID=1068.2.0
BUILD_ID=2016-06-17-1616
PRETTY_NAME="CoreOS 1068.2.0 (MoreOS)"
ANSI_COLOR="1;32"
HOME_URL="https://coreos.com/"
BUG_REPORT_URL="https://github.com/coreos/bugs/issues"

Environment

VMware ESXi

Expected Behavior

Interface to become configured and systemd-networkd-wait-online not to fail.

Actual Behavior

Interface never gets configured state:

$ networkctl status ens192
● 2: ens192
       Link File: /usr/lib64/systemd/network/99-default.link
    Network File: /etc/systemd/network/static.network
            Type: ether
           State: routable (configuring)
            Path: pci-0000:0b:00.0
          Driver: vmxnet3
          Vendor: VMware
           Model: VMXNET3 Ethernet Controller
      HW Address: 00:50:56:xx:xx:xx (VMware, Inc.)
             MTU: 1500
         Address: xxx.xx.xx.xxx
         Gateway: xxx.xx.xx.xxx (Check Point Software Technologies)
             DNS: xxx.xx.xx.xxx
                  xxx.xx.xx.xxx

And systemd-networkd-wait-online fails:

$ systemctl status systemd-networkd-wait-online
● systemd-networkd-wait-online.service - Wait for Network to be Configured
   Loaded: loaded (/usr/lib64/systemd/system/systemd-networkd-wait-online.service; disabled; vendor preset: disabled)
   Active: failed (Result: exit-code) since Wed 2016-06-22 11:26:44 UTC; 41min ago
     Docs: man:systemd-networkd-wait-online.service(8)
 Main PID: 749 (code=exited, status=1/FAILURE)

Jun 22 11:25:06 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:25:06 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:25:07 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:25:09 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:25:09 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:25:10 kl1258pm.is.klmcorp.net systemd-networkd-wait-online[749]: ignoring: lo
Jun 22 11:26:44 kl1258pm.is.klmcorp.net systemd[1]: systemd-networkd-wait-online.service: Main process exited, code=exited, status=1/FAILURE
Jun 22 11:26:44 kl1258pm.is.klmcorp.net systemd[1]: Failed to start Wait for Network to be Configured.
Jun 22 11:26:44 kl1258pm.is.klmcorp.net systemd[1]: systemd-networkd-wait-online.service: Unit entered failed state.
Jun 22 11:26:44 kl1258pm.is.klmcorp.net systemd[1]: systemd-networkd-wait-online.service: Failed with result 'exit-code'.

Reproduction Steps

1. Configure interface statically
2. Disable IPv6 with above drop-in (net.ipv6.conf.all.disable_ipv6)
3. Boot machine
4. networkctl status, will show interface as configuring

Other Information

I've not been able to test if this also happens when using DHCP.

[mischief]

what happens if you set:

[Network]
LinkLocalAddressing=no

[basvdlei]

Nope, that doesn't change the behavior unfortunately. In the same vein, I also tried adding DHCP=no, which also didn't help.

[mischief]

i checked and confirmed this happens in QEMU.

[mischief]

already reported upstream at systemd/systemd#2713

[qrpike]

Any updates on this? I can't upgrade CoreOS because of this bug...

[basvdlei]

The behavior has changed in the current beta (for the worse), because networkd actually enables IPv6...

Here is an example of a stable and beta machine, both having a sysctl dropin setting net.ipv6.conf.all.disable_ipv6 = 1:

CoreOS 1122.2.0 (systemd 229):

sudo sysctl -qa 2>/dev/null  | grep 'disable_ipv6'
net.ipv6.conf.all.disable_ipv6 = 1
net.ipv6.conf.default.disable_ipv6 = 1
net.ipv6.conf.docker0.disable_ipv6 = 1
net.ipv6.conf.eth0.disable_ipv6 = 1
net.ipv6.conf.lo.disable_ipv6 = 1

CoreOS 1153.4.0 (systemd 231):

sudo sysctl -qa 2>/dev/null  | grep 'disable_ipv6'
net.ipv6.conf.all.disable_ipv6 = 1
net.ipv6.conf.default.disable_ipv6 = 1
net.ipv6.conf.docker0.disable_ipv6 = 0
net.ipv6.conf.eth0.disable_ipv6 = 0
net.ipv6.conf.lo.disable_ipv6 = 1

[dm0-]

The issue is that systemd's system-wide IPv6 detection function is broken. It looks like you can add both of these lines to your [Network] section to work around the problem for now, which make networkd think IPv6 is disabled for the matching interfaces.

LinkLocalAddressing=no
IPv6AcceptRA=no

A fix should be included in the next alpha that will work with the .network file from your configuration as-is.

[dm0-]

Upstream systemd has rejected fixing this. The recommended solution is to write your network files to disable IPv6 as described in the previous comment.