I would chime in a bit on this too. 14.04 _did_ have such issues. In
fact, I had a physical box with static IP's (2 nics), if a cable was
unpligged, it would hang forever and also become inconsistent. I had
these discussions a year ago or so. Anyway, I had ot abandon ubuntu
server (along with some nft issues at the time) and run archlinux as my
But I ran zfs as my xen backend, so on arch, I sometimes* could have
issues ~ once a year where it best be I had physical access to the box
to be able to fix any potential issues.
I am now going to move abroad and so I thought , with archlinux unpriv.
lxc containers out, to leave xen (as I don't have the expected network
output from the servers) to lxc. With arch and benig abroad, I think
ubuntu 16.04 now with systemd would have solved this (my server with
arch on xen has always gracefully sorted it out: ~1.5-2 years).
Seeing this bug and seein git has to do with ifup makes me think it's
partly a remnance from the older logic from sysv/upstart.
This probably doesn't help much butjust saying the ifup is a bit archaic
now that systemd is in use and is not evebn present in archlinux and
creates seemingly less problems.
I have only a brief moment before leaving so seeing this now worries me.
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to ifupdown in Ubuntu.
ifup does not block for interface to be up with static addresses
Status in ifupdown package in Ubuntu:
Description: Ubuntu 16.04 LTS (server)
*** 229-4ubuntu6 500
500 http://us.archive.ubuntu.com/ubuntu xenial-updates/main amd64
500 http://us.archive.ubuntu.com/ubuntu xenial/main amd64 Packages
This issue is more about systemd integration then an issue with
systemd package itself. The general problem is the distinction
between the network being configured and the network being up and what
that means for services that depend on the network.
upstream systemd has network.target (network is configured) and
network-online.target (network is up), however xenial server (dunno
about desktop) doesn't seem to use network-online.target. The result
of which is, as soon as the network is configured both network.target
and network-online.target are considered complete. At which point
services dependent on either of those will get started. However this
can lead to problems if the services in question are actually
dependent on the network (ie dns lookups) when they are starting up.
Here's an example of what I mean. If you install openntpd and add the
-s option there is a race on boot if it will or wont be able to do the
Jun 02 16:58:02 xenial64 systemd: Reached target Network.
Jun 02 16:58:02 xenial64 systemd: Starting OpenNTPd Network Time
Jun 02 16:58:02 xenial64 systemd: Reached target Network is Online.
Jun 02 16:58:02 xenial64 ntpd: constraint certificate verification
Jun 02 16:58:02 xenial64 ntpd: ntp engine ready
Jun 02 16:58:02 xenial64 kernel: tg3 0000:02:00.0 enp2s0: Link is down
Jun 02 16:58:05 xenial64 kernel: tg3 0000:02:00.0 enp2s0: Link is up at 1000
Mbps, full duplex
Jun 02 16:58:05 xenial64 kernel: tg3 0000:02:00.0 enp2s0: Flow control is on
for TX and on for RX
Jun 02 16:58:05 xenial64 kernel: IPv6: ADDRCONF(NETDEV_CHANGE): enp2s0: link
Jun 02 16:58:17 xenial64 ntpd: no reply received in time, skipping
initial time setting
openntpd was started 3 seconds before the interface even got link.
I believe this issue is only going to effect physical servers with
static ips. vm's interfaces will instantly have link when they are
brought up and dhcp, the interface will already be up or else they
wouldn't have been able to get an ip.
To manage notifications about this bug go to:
Mailing list: https://launchpad.net/~touch-packages
Post to : email@example.com
Unsubscribe : https://launchpad.net/~touch-packages
More help : https://help.launchpad.net/ListHelp