[GE users] Default queue state for rebooted machines?

Fred L Youhanaie fly at anydata.co.uk
Wed May 4 14:39:54 BST 2005

    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi Rob,

(strangely your email is from 11:26, but I've only just received it from 

>>Without attempting to trivialize your application, going for this 
>>solution would be a matter of putting 'touch /tmp/shared-data-ready' at 
>>the end of the shell script - assuming that the shared data is ready at 
>>the end of the script.
> I can see how this would work, given that /tmp is local to each machine, but
> this presumes that /tmp is cleaned out during a shutdown/reboot procedure.
> Is this the case? Even in "spontaneous" reboots without apparent cause?
> However, this method of checking for the existence of this file would still
> allow jobs to be scheduled to this machine and we'd still have the
> "black-hole" problem that Magnus mentioned! 

You are right, I did not provide a complete solution, somehow you do 
need to ensure that the file is not there when the data is not ready 
yet! Typically /tmp under solaris is on swap space, but for linux you 
would need to create such a directory. Even with such safeguards as soon 
as the shared data becomes invalid (not due to reboots) you end up with 
the black-hole problem.

Reuti's suggestion of using the prolog scripts seems good. Is your 
application developed in house? If so, can the developers provide you 
with a small program (or code snippet for you to compile) such that it 
checks the shared data segment and returns Yes/No answer? You can then 
put run this program in the prolog. Remember, you can access the shm via 
perl :)


To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

More information about the gridengine-users mailing list