[Condor-users] job restart delay after system crash

[Horvátth Szabolcs <szabolcs@xxxxxxxxxxxxx>]

Hi,

We had some very similar system crashes lately (because of disk failures 
of a specific type of hardware)
and condor responded much the same way every time: After the computer 
that runs the starter dies, condor_q
still reports that it is running on that machine for eternity (at least 
more than 6 hours), even though condor_starter
does not show that computer at all.

The only way I found to remove the job from that crashed machine is to 
hold and release the task (condor_vacate and
condor_vacate_job does not work) but even after release the job does not 
get restarted and stays in the idle state for hours.

I'm not really sure whether its simply a configuration issue or a bug so 
first I'd like to ask which configuration settings
might affect this behaviour? How can I make condor realize that the 
scheduler is not getting any response from the starter
and after that to re-negotiate the job instead of waiting for hours?

(Using condor 6.8.4 on window XP.)

Cheers,
Szabolcs