[Techtalk] Basic connection questions

[Wim De Smet]

On Tue, May 28, 2013 at 6:50 PM, vtrucs <vtrucs at gmail.com> wrote:

> Wim De Smet wrote:
>
>> FWIW, NetworkManager logs most of what it does to syslog
>> (/var/log/syslog).
>> So if you read the stuff that gets logged right after boot, you should be
>> able to see if it is trying to manage your network interface or not.
>>
>
> I have been checking syslog, as well as daemon-log, messages and dmesg.
> Daemon-log merely repeats the dhcpdiscover efforts with no further
> information. Syslog repeats those too, but has this interesting message:
> May 28 12:13:06 none NetworkManager: <info> Unmanaged Device found; state
> CONNECTED forced. (see http://bugs.launchpad.net/**bugs/191889<http://bugs.launchpad.net/bugs/191889>
> )
> as does dmesg when it first tries to bring up the connection.
>
> However, the bug listed seems to have more to do with a browser always
> coming up off-line, which is not a problem for me. Also, the Network
> Manager icon show eth0, the wired connection, as "device not managed."


That sounds good. It shouldn't be doing anything to that interface then.


>
>
>   I assume you already configured it to run in unmanaged mode.
>>
>>
>>
>
> The only way I know to do this is to list the device in
> /etc/network/interfaces. Listing/not listing it seems to be what toggles
> the managed/not managed message in the NM icon. NM's conf file has
> managed=false for the [ifupdown], that and keyfile are the only plugins
> listed under [main] and the no-auto-default is set to the mac address of
> the wired LAN card. H'mmm.


no-auto-default is not necessary as far as I know, but that sounds like
it's correctly configured.


>
>  To me it sounds like this is not an NM issue at all, but probably more of
>> a
>> DHCP configuration issue. For instance, some servers require you to
>> release
>> your previous lease if you do a fresh dhcprequest without releasing the
>> previous one, even if the standard probably says they should just reissue
>> it.
>>
>>
>
>>
>
> I noticed that when I run ifdown on the device, it does do a DHCPRELEASE
> on an IP address, and if I run dhclient, a message that an alreadly running
> process ID had been killed. However, the problem still happens even with
> the release and PID kill. This is interesting, though - I had not tried to
> connect eht0 since logging in today, so it appears something else did.
>

At boot, ifup should be run, so that would start the original dhclient
process that's trying to get an IP onto the interface. I don't know why it
wouldn't be getting an IP this way, normally dhclient works even if NM is
managing the interface. Do you have any sort of firewalling configured?

I am thinking it might have something to do with the dhcp server since I've
spent far too much time trying to debug DHCP issues only to have the
problem go away on rebooting or resetting whatever was supposed to function
as a DHCP server.

Wim