Home / ubuntu / Questions / 833276
Accepted
Ray
Asked: 2016-10-06 00:15:20 +0800 CST

Network connection lost ('carrier-changed')

  • 772

I'm having a problem with wired network connections of Ubuntu systems (both 15.10 and 16.04). Basically, connection is suddenly lost.

I'm trying to figure out if it's a problem with:

  • network cable,
  • switch,
  • Ubuntu, or
  • something else

I actually doubt it is related to Ubuntu (since the problem is with multiple computers) but am posting here in the hopes that someone can suggest what to look for in the log files.

The problem I've seen is with two network switches, both of which are new and are of the same make and model.

syslog indicates the date and time of the problem, but I don't really know how to interpret it. For example, this is in the log:

Oct  4 18:30:10 ... NetworkManager[1695]: <info>  [1475577010.9502] device (eno1): link disconnected (calling deferred action)
Oct  4 18:30:10 ... NetworkManager[1695]: <info>  [1475577010.9504] device (eno1): state change: activated -> unavailable (reason 'carrier-changed') [100 20 40]

Is there a list of "state changes" for NetworkManager which could help?

For example, I was playing around with another computer and if I switched off the network using software, I get:

unavailable -> disconnected

Whereas if I unplugged it, I get:

activated -> unavailable

But this is just a guess... I've reached the point where I really don't know what I'm doing.

Any suggestions would be appreciated! Thank you!

Ray

networking

3 Answers

  1. Best Answer

    After removing other factors, such as the switch, the problem continued.

    I asked one of the system administrators of our organisation's network and he seems to think that such an error message means that there is a physical problem with the network.

    For example, with the switch gone, this means:

    1. cable from the computer to the port
    2. port itself
    3. even the cabling behind the port, within the wall

    In any case, he said it is "definitely" not related to the OS but was unable to decipher the error message piece-by-piece. So, this is where the problem stands and I'll be looking at the physical connection to the computer instead of Ubuntu.

    • 2

  2. In my case carrier-changed coincides with the router disconnecting/reconnecting with the internet (see the logs in your router). The cause may be a not well connected cable, but there are several other possibilities: cable provider initiates a reconnect; a router may have problems (firmware; data backup loaded from an other router where the data somehow confuses this router ...).

    However, the computer should successfully reconnect to the router itself when the router has its uplink working again. In my case the router needed 12 minutes (!) with several "cable internet synchronising (training)" attempts before establishing a stable uplink. Maybe the cable provider installs several firmware updates remotely ( I can't even see what firmware is running at the router (fritzbox 6490), the cable provider has full control - annoying!) - but this is just speculation.

    I forced a reconnect of the router myself which took only 7 secs. The computer has had no problem reconnecting in this case. I assume this is because of the short time frame.

    Question remains: how to bring the computer to self heal a carrier-change even if this change takes a lot of time.

    • 1

  3. You can try what is suggested in NetworkManager does not react to link state changes in RHEL7, 8 Server:

    Environment

    • Red Hat Enterprise Linux 7
    • Red Hat Enterprise Linux 8
    • NetworkManager

    Issue

    • When a link goes down, NetworkManager does not remove its IP address. This means any routes associated with the interface are still present in the kernel routing table and connectivity over backup interfaces/gateways does not work as expected.
    • This is different than the behaviour of NetworkManager in RHEL 6.

    Resolution

    • Disable the NetworkManager ignore-carrier option:

      • Create a file in the /etc/NetworkManager/conf.d/ directory beginning with a number greater than 00, such as 99. with the following contents:

        # cat /etc/NetworkManager/conf.d/99-carrier.conf
[main]
ignore-carrier=no

      • Restart the NetworkManager service:

        # systemctl restart NetworkManager

    Root Cause

    • By default, RHEL7 and RHEL8 Server includes the NetworkManager-config-server package. This package modifies the default NetworkManager behaviour to:

      1. Disable auto-configuration of all interfaces by enabling the no-auto-default option
      2. Disable monitoring of link state by enabling the ignore-carrier option
    • -3