[Mageia-dev] systemd network.service issues

[Anssi Hannula]

16.03.2012 14:45, Colin Guthrie kirjoitti:
> 'Twas brillig, and Anssi Hannula at 16/03/12 03:45 did gyre and gimble:
>> Hi all!
>>
>> So, I've just upgraded my home server to mga2/cauldron.
>>
>> Something seems fishy in the network.service handling, systemd thinks it
>> has failed:
>>
>> # systemctl status network.service
>> network.service - LSB: Bring up/down networking
>>           Loaded: loaded (/etc/rc.d/init.d/network)
>>           Active: failed (Result: exit-code) since Fri, 16 Mar 2012
>> 05:25:09 +0200; 11min ago
>>           CGroup: name=systemd:/system/network.service
>>                   ├ 4679 /sbin/ifplugd -I -b -i eth0
>>                   ├ 4723 /sbin/ifplugd -I -b -i eth1
>>                   └ 5372 /sbin/dhclient -q -lf
>> /var/lib/dhcp/dhclient-eth0.leases -pf /var/run/dhclient-eth0.pid -cf
>> /etc/dhclient-eth0.conf...
>>
>> Mar 16 05:25:09 delta.onse.fi ifplugd(eth1)[4723]: Using detection mode:
>> SIOCETHTOOL
>> Mar 16 05:25:09 delta.onse.fi ifplugd(eth1)[4723]: Initialization
>> complete, link beat not detected.
>> Mar 16 05:25:09 delta.onse.fi network[4397]: [161B blob data]
>> Mar 16 05:25:09 delta.onse.fi network[4397]: [14B blob data]
>> Mar 16 05:25:13 delta.onse.fi ifplugd(eth0)[4679]: Link beat detected.
>> Mar 16 05:25:14 delta.onse.fi ifplugd(eth0)[4679]: Executing
>> '/etc/ifplugd/ifplugd.action eth0 up'.
>> Mar 16 05:25:15 delta.onse.fi dhclient[5312]: DHCPREQUEST on eth0 to
>> 255.255.255.255 port 67
>> Mar 16 05:25:15 delta.onse.fi dhclient[5312]: DHCPACK from 109.204.128.1
>> Mar 16 05:25:15 delta.onse.fi ifplugd(eth0)[4679]: client: Determining
>> IP information for eth0... done.
>> Mar 16 05:25:15 delta.onse.fi ifplugd(eth0)[4679]: Program executed
>> successfully.
>>
>>
>> Trying to restart it I seem to hit other issues:
>>
>> # systemctl restart network.service
>> Job failed. See system journal and 'systemctl status' for details.
>> # systemctl status network.service
>> network.service - LSB: Bring up/down networking
>>           Loaded: loaded (/etc/rc.d/init.d/network)
>>           Active: failed (Result: exit-code) since Fri, 16 Mar 2012
>> 05:36:39 +0200; 3s ago
>>          Process: 8027 ExecStart=/etc/rc.d/init.d/network start
>> (code=exited, status=1/FAILURE)
>>           CGroup: name=systemd:/system/network.service
>>                   ├ 8195 /sbin/ifplugd -I -b -i eth0
>>                   └ 8353 /sbin/dhclient -q -lf
>> /var/lib/dhcp/dhclient-eth0.leases -pf /var/run/dhclient-eth0.pid -cf
>> /etc/dhclient-eth0.conf...
>>
>> Mar 16 05:36:39 delta.onse.fi network[8027]: RTNETLINK answers: File exists
>> Mar 16 05:36:39 delta.onse.fi network[8027]: RTNETLINK answers: File exists
>> Mar 16 05:36:39 delta.onse.fi network[8027]: RTNETLINK answers: File exists
>> Mar 16 05:36:39 delta.onse.fi network[8027]: RTNETLINK answers: File exists
>> Mar 16 05:36:39 delta.onse.fi dhclient[8261]: DHCPREQUEST on eth0 to
>> 255.255.255.255 port 67
>> Mar 16 05:36:40 delta.onse.fi dhclient[8261]: DHCPACK from 109.204.128.1
>> Mar 16 05:36:40 delta.onse.fi ifplugd(eth0)[8195]: client: Determining
>> IP information for eth0...NETLINK: Error: File exists
>> Mar 16 05:36:40 delta.onse.fi ifplugd(eth0)[8195]: client:  done.
>> Mar 16 05:36:40 delta.onse.fi ifplugd(eth0)[8195]: client: NETLINK:
>> Error: File exists
>> Mar 16 05:36:40 delta.onse.fi ifplugd(eth0)[8195]: Program executed
>> successfully.
>>
>> Any ideas? If not, I'll investigate this myself later.
> 
> network.service fails here too, but I've not been able to work out
> exactly why. It doens't bother me too much tho' as all my devices are
> using network manager!
> 
> THat said, it seems that network.service has still started an ifplugd
> for my wlan, despite the ifcfg file clearly saying that NM should take
> care of that interface... needs some investigating for sure.
> 
> 
>> On a related note, I also got dropped to emergency console on first
>> boot, not sure why was that (it didn't tell me the reason... maybe the
>> reason was in another virtual console, but I didn't remember to check
>> there at the time). I just exited the console and it continued to boot...
> 
> Interesting. Was it a dracut shell or a systemd one? (sometimes hard to
> tell which is which, but the text usually tells you.

I think it was a systemd one (it seemed to be quite late in the boot,
when I exited it the next thing it did was build DKMS stuff).

> Knowing which should help identify why. Does it happen on subsequent boots?

No idea yet, but will reboot the system now.

>> I see the following in systemctl --failed:
>>
>> # systemctl --failed
>> UNIT                         LOAD   ACTIVE SUB    JOB DESCRIPTION
>> bootparamd.service           loaded ESC[1;31mfailed failedESC[0m
>> SYSV: The bootparamd server allows older Sun workstations to net boot
>> from Linux boxes. It (along with rarp) is rarely used anymore; bootp and
>> dhcp have mostly replaced both of them.
>> fedora-loadmodules.service   loaded ESC[1;31mfailed failedESC[0m
>> Load legacy module configuration
> 
> dmorgan reported this as well, but it's a really simple script and
> really shouldn't fail.
> 
> I think it must be the /etc/rc.modules script that is likely failing
> here. Perhaps trying to modprobe a module that no longer exists due to
> old configuration? If possible can you debug the fedora-load-modules
> service and comment out one half of the script such that only the
> rc.modules bit is run and confirm that's where it's bust. Then maybe
> work out what in rc.modules is broken. That woudl be useful as it works
> fine here.

Actually it is because /etc/modprobe.preload contains some non-existing
modules (like psmouse), and /etc/modprobe.preload.d/cpufreq
(cpufreq_performance).

bootparamd failed because I had no /etc/bootparamd. I'm pretty sure I've
not installed bootparamd myself, but it was a leaf package now so I
simply removed it.

>> network.service              loaded ESC[1;31mfailed failedESC[0m
>> LSB: Bring up/down networking
> 
> Mentioned above. Needs to be debugged.
> 
>> systemd-modules-load.service loaded ESC[1;31mfailed failedESC[0m
>> Load Kernel Modules
> 
> Yeah this one fails for me too. Not sure why, but maybe because I've
> done something stupid and symlinked stuff into /etc/modules-load.d when
> I shouldn't. I'll look into it.

I guess this would be for the same reason as fedora-loadmodules.service...

BTW, isn't this duplication? Well, I guess not completely,
/etc/modules-load.d doesn't contain the stuff from
/etc/modprobe.preload.d/, only /etc/modprobe.preload and /etc/modules.

> 
> Incidentally, just answering your private tmp question from IRC...
> 
> This is what Lennart had to say about it:
> <mezcalero> coling: the /tmp dir used by PrivateTmp lies beneath /tmp
> <mezcalero> coling: hence, if /tmp is cleaned up those private tmps will
> be cleaned up too
> <mezcalero> coling: since the clean-up is recursive

Ah, clever. Though I should've recognized it was just a simple
bind-mount... I'll blame my tiredness for this as well :)

-- 
Anssi Hannula