0

my Ubuntu 20.04 LTS server looses its inbound LAN connection after some days of operation. The outbound connection works: I receive status e-Mails from that host. Bridged containers work as they should.

I have seen the resolved answer about a 20.04 LTS, where the network is unreachable. Basically it seems like I have the same problem, since I am not that firm with netplan, maybe someone can point me into the right direction?

My configuration under /etc/netplan/bridge.yaml looks like this:

network:
  version: 2
  renderer: networkd
  ethernets:
    enp0s25:
      dhcp4: yes
      dhcp6: yes
  bridges:
    br0:
      dhcp4: yes
      dhcp6: yes
      interfaces:
        - enp0s25

The hardware has two devices enp0s25 and enp5s0. I once tried to bond them and later thought about separating them (regular uplink via enp0s25 and bridge via enp5s0) but that did not work.. In my current configuration I simply deactivated the second device.

However, what puzzles me now is, that the host itself not reachable, but everything bridged works fluently.

Thanks in advance, Dorian

Update

We have restarted the server, so that I can access it remotely. Thereby, a colleague recognized, that the IPv6 address still works. I have run journalctl -S 2020-12-18 -o short-full -p 3 to see the error events since the last boot and found a lot of systemd-networkd-wait-online[2579361]: Event loop failed: Connection timed out events. → I'm investigating

Dorian
  • 1
  • On the server, run ip addr. Is the IP address of the server the same address that you expect? – user535733 Jan 22 '21 at 12:54
  • I can't check right now, because... we once again lost our connection :) The IP is static, though. And it works in the beginning, there should be no reason for a DHCP lease. – Dorian Jan 22 '21 at 14:28
  • Your netplan specifies dhcp "yes", which implies that you are using a router setting for the IP address. Login to the router and check that setting. – user535733 Jan 22 '21 at 16:05
  • Yea, that is what I meant.The listing of network devices is as it should be. The right IP is assigned to the host. Still, when I ping it via IP or hostname the request times out. – Dorian Jan 22 '21 at 16:42
  • If you try to ping or ssh the server, what's the actual, complete error message? – user535733 Jan 22 '21 at 16:49
  • It is the simple Request timeout for icmp_seq 0 and in case of SSH basically the same ...connect: operation timed out.

    It behaves like the physical connection was severed or as if the firewall rules were altered. But in fact the basic configuration worked for years and through 16.04 LTS and 18.04 LTS .. only since 20.04 and the adaptation of netplan, everything went downhill.

    The connection is lost after several days of operation. Sometimes it reapers, often times it does not until a reboot.

     – Dorian Jan 22 '21 at 17:21

1 Answers1

0

Well, in my despair I wanted to clean install the device, which would have been a lot of trouble due to migrations.

So, I still wanted to try static IPs via netplan and in search for the correct syntax stumbled upon this article . It basically says to remove the ifupdown package (apt remove ifupdown), which I was not even aware, that an distupgrade would still keep remnants.

However, either the static IP or the removal of ifupdown seems to solve the issue.

Thanks @all.

Dorian
  • 1