[gridengine users] Use a queue for redundancy
reuti at staff.uni-marburg.de
Thu Jun 21 11:48:55 UTC 2012
Am 21.06.2012 um 11:35 schrieb Joe Borġ:
> It's not just SGE that run on these, they are designed to fail over. It's critical stuff on them, so not really a waste.
> Why when I set "Slots" to 1 can a job start on each node? Is this the way it's mean to work?
Yes, it's the slot count per queue instance - i.e. per exechost in the defined hostlist.
But using a consumable complex might also not work, as it's still marked as used in SGE's internal bookkeeping I would say. A vanishing node shouldn't change the complex consumption for already running jobs.
Therefore the idea with the load sensor.
> Joseph David Borġ
> On 20 June 2012 17:41, Reuti <reuti at staff.uni-marburg.de> wrote:
> Am 20.06.2012 um 18:22 schrieb Joe Borġ:
> > I've got a queue with two hosts in. I'd like to set this up so host2 is only used when host1 goes into state "au". Currently I'm using consumable resources to limit one job at a time, but this doesn't seem the best way.
> You mean host2 is idle under normal circumstance - isn't it a waste of resources when the machine is running anyway? You could use a load sensor on host2 which checks the availability of host1 and enables/disables the queue instance on host2.(the load sensor could change a boolean complex).
> -- Reuti
> > Any suggestions?
> > Regards,
> > Joseph David Borġ
> > http://www.jdborg.com
> > _______________________________________________
> > users mailing list
> > users at gridengine.org
> > https://gridengine.org/mailman/listinfo/users
More information about the users