Re: [HTCondor-users] hostname empty on node restart

[Yngve Inntjore Levinsen <yngve.levinsen@xxxxxxx>]

Hi Rich,

> while [ `hostname` = localhost ]; do sleep 5; done
Simple and excellent idea, thanks!

What I did meanwhile, which seems to work well so far, is that I created 
a new init script. I have removed condor from all run levels, adding 
this new script instead with a very simple start() command:

start() {
        echo "Delayed Condor server startup"
        sleep 30
        service condor start
        echo
}

This delay is a bit longer than you suggest, but it doesn't matter to us 
if we lose 30 seconds of simulation time after a startup. It also avoids 
the while loop (e.g. if network cable disconnected?). If anyone would 
manually run 'service condor start' they would not be affected by this 
delay.

Not sure if I will run into other surprising issues as a result of this 
change, but so far so good.

Cheers,
Yngve


On 04/06/15 15:53, Rich Pieri wrote:
> On 6/4/15 4:28 AM, Yngve Inntjore Levinsen wrote:
> >  From what I understand about run levels, I thought that if network is
> > needed then it shouldn't go in run level 2, only 3-5. This seems to not
> > be the case for these machines though, where level 2 is the default
> > after startup.
> Only run levels 0, 1 (or S) and 6 are set in stone. LSB suggests that
> level 2 is for full multi-user without networking and level 3 is level 2
> plus networking. You should still look at your distribution target
> because many long-standing distributions like Debian and Slackware
> predate LSB.
>
>
> > The change seems to work half way. Sometimes the machines get the
> > correct network name, sometimes they are displayed as 'localhost', and
> > sometimes with empty hostname.
> This may be due to latency in the DHCP requests: dhcpcd is running in
> the background waiting for a lease when the Condor daemon start happens.
> If this is the case then you can't fix it by changing start priorities.
> You need to make Condor's start process dependent on DHCP lease
> acquisition. A simple hack is a little semi-infinite loop in the init
> script:
>
> while [ `hostname` = localhost ]; do sleep 5; done