Systemd on vivid beta

[Colin Law]

On 12 March 2015 at 13:30, Tom H <tomh0665 at gmail.com> wrote:
> On Thu, Mar 12, 2015 at 7:51 AM, Colin Law <clanlaw at gmail.com> wrote:
>>
>> The plot thickens!
>> Having worked around the fact that syslog closes down too soon by
>> adding diy logging to a file from the script I find that the culprit
>> is a line of the form
>> umount -l /path/to/mountpoint
>> where this was orginally mounted using something like
>> mount -t cifs -o credentials=***,user,rw,uid=1000 //192.168.1.nnn/path
>> /path/to/mountpoint
>>
>> The umount command takes anything up to 2 minutes to complete.
>> If I use systemctl stop to stop the service it runs straight through
>> without delay.  I wonder whether some networking service has also been
>> stopped too soon and umount is timing out waiting for a response.
>
> I agree with the "network's gone down before the unmount" theory. I'm
> pretty sure that I've hit this problem with an nfs mount.
>
> If you're using NM, there's just been an update to adjust the
> dependencies of NetworkManager-wait-online.service and enable it that
> might resolve your problem.

Thanks for the suggestion, I am using NM, but enabling that service
did not appear to make any difference.  Looking at the information I
could find on that it seemed that the effect is the same as including
After=network-online.target and Wants=network-online.target in the
service file, and they are there already.

However I think maybe the problem is more fundamental than I thought.
I find that if I just manually mount a cifs share ( sudo mount -t cifs
...) and shutdown that I get the same delay on shutdown, whereas if I
do the same in an upstart boot then there is no delay.  That seems
like a reportable bug to me, so I have reported it.
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1431774

Colin