Bug#791391: NFS mounts and other things fail if DHCP hasn't finished

Sat Jul 4 17:08:06 BST 2015

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 04/07/15 14:29, Michael Biebl wrote:
> Am 04.07.2015 um 14:22 schrieb Daniel Pocock:
>> 
>> 
>> On 04/07/15 13:58, Michael Biebl wrote:
>>> Am 04.07.2015 um 11:38 schrieb Daniel Pocock:
>>>> Package: systemd Version: 215-17+deb8u1 Severity: important
>>>> 
>>>> This has been happening on a system that was upgraded from
>>>> wheezy to jessie
>>>> 
>>>> Sometimes systemd starts the network and then tries to start 
>>>> things that depend on the network (e.g. ntpdate, NFS client
> 
> Isn't ntpdate (*) started via an if-up.d hook i.e. not by systemd
> itself?
> 

Yes - I see that I have /etc/network/if-up.d/ntpdate

grep ntpdate /etc/init.d/* doesn't find anything.  It is not run from
cron or rc.local either.

Looking at journalctl output, it looks like it tried to run ntpdate
twice.  I've copied the relevant events below:

Jul 04 11:26:19 ntpdate[1182]: Can't find host ntp: Name or service
not known (-2)
Jul 04 11:26:19 ntpdate[1182]: no servers can be used, exiting

Jul 04 11:26:19 daniel1 kernel: tg3 0000:02:00.0: irq 105 for MSI/MSI-X
Jul 04 11:26:19 daniel1 kernel: IPv6: ADDRCONF(NETDEV_UP): eth0: link
is not ready

Jul 04 11:26:19 daniel1 networking[1111]: Configuring network
interfaces...done.

Jul 04 11:26:19 daniel1 dhclient[1221]: Internet Systems Consortium
DHCP Client 4.3.1
Jul 04 11:26:20 daniel1 dhclient[1221]: DHCPDISCOVER on eth0 to
255.255.255.255 port 67 interval 6

Jul 04 11:26:20 mount[1268]: mount.nfs: Network is unreachable

Jul 04 11:26:23 ntpd[1596]: ntpd 4.2.6p5 at 1.2349-o Fri Apr 10 19:04:04
UTC 2015 (1)
Jul 04 11:26:23 ntp[1564]: Starting NTP server: ntpd.

Jul 04 11:26:26 ifup[1199]: DHCPACK from 192.168.1.2
Jul 04 11:26:26 NetworkManager[1433]: <info> (eth1): preparing device
Jul 04 11:26:26 NetworkManager[1433]: <info> (eth1): created default
wired connection 'Wired connection 1'
Jul 04 11:26:26 NetworkManager[1433]: <info> NetworkManager state is
now CONNECTED_GLOBAL

Jul 04 11:26:29 ntpdate[2157]: the NTP socket is in use, exiting

> 
>>>> mounts) before DHCP has obtained a lease.
>>>> 
>>>> Sometimes the boot completely stops and asks for root login
>>>> 
>>>> Sometimes it reaches the X login but if a user logs in, it
>>>> throws them out because their home is not mounted.
>> 
>>> Is home mounted via NFS?
>> 
>> 
>> /home is a local filesystem
> 
> Is /home mountend reliably by systemd?
> 

Actually, /home is part of the root filesystem, it is not a separate
filesystem.  It always seems to mount the root filesystem
successfully.  Sometimes I have had systemd grumble trying to mount
local filesystems twice on another machine but I haven't seen that
happen on this one.

>> /home/username is on NFS
> 
> How is /home/username mounted? Via /etc/fstab, autofs, something
> else?

Just /etc/fstab - it is very basic:

192.168.1.2:/home/daniel /home/daniel nfs rw,vers=3 0 2

> 
> Just trying to find out what exactly fails and what not.
> 
> 
> (*) As a side note: you might want to consider enabling 
> systemd-timesyncd.service (and removing ntpdate). timesyncd is a
> very lightweight NTP client.
> 

Would this mean removing both the ntp and ntpdate packages?  They are
all configured to talk to an NTP server on 192.168.1.2