As pointed out by Lucasz, fork jobs run directly on the node on which GRAM
services (head-node)are deployed and not on the compute nodes (where
scheduled jobs run). This behavior is often useful when you have your
headnode doubling up as a compile node. If for some reason, I would want
my code to say compile, only on the headnode and not on a compute node, I
would then submit the job to the fork jobmanager.
Regards,
Prashanth Chengi
National PARAM SuperComputing Facility
System Administration and Networking Group
C-DAC Pune
Phone: 02025704197
--
Now we that are strong ought to bear the infirmities of the weak,
and not please ourselves.
Romans 15:1
On Thu, 13 May 2010, Lukasz Lacinski wrote:
Exactly. Jobs submitted to the fork job manager are executed immediately. For
example when you submit quickly five jobs taking 60 secs (let's say `/bin/sleep
60`) all of these jobs will be executed immediately on the node with GRAM
service, run simultaneously and share resources for about 60 secs.
Lukasz
On May 13, 2010, at 3:00 PM, [email protected] wrote:
Hi
Can someone tell me what the charakteristics of the fork jobmanager are? I read in a
description "unscheduled local execution" does that mean that this jobmanager
has no scheduling function? just immediate execution of the job - something like FIFO?
thx
--
GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
