Systemd service life cycle

Tom H tomh0665 at gmail.com
Sun Apr 10 11:00:39 UTC 2016


On Sat, Apr 9, 2016 at 11:06 PM, Colin Law <clanlaw at gmail.com> wrote:
> On 9 April 2016 at 21:59, Tom H <tomh0665 at gmail.com> wrote:
>> On Sat, Apr 9, 2016 at 6:20 PM, Colin Law <clanlaw at gmail.com> wrote:
>>> On 9 April 2016 at 16:44, Tom H <tomh0665 at gmail.com> wrote:
>>>>
>>>> "oneshot" is better for your use-case. I forgot...
>>>
>>> What is the reason for oneshot being better than forking in this case?
>>>
>>> I need to leave the tunnel process running so I would have thought
>>> forking is a better fit.
>>>
>>> In fact this has all become academic (though still interesting) as I
>>> realised that actually I would be much better to move the process at
>>> the other end of the tunnel into this server and then I don't need the
>>> tunnel, and in fact it is better there for other reasons anyway.
>>
>> It's for a unit that doesn't have any active processes after ExecStart
>> runs. I've used it for setting up networking and firewalling, for
>> example.
>
> Is not the tunnel an 'active process' though?

Only if there's a daemon that's maintaining it up.

For example, when I run "iptables-restore ..." from a systemd unit, it
loads the iptables rules and exits.




More information about the ubuntu-users mailing list