1)Interaction
Method.
2)Via
script we can submit the Job.
1)Interaction
Method
#qsub
-l select=1:ncpus=2+1:ncpus=1 – (Command (or) Job-Script)
|
2)Script
Method
$ cat
script.sh
#!
/bin/bash
#PBS -l
select=1:ncpus=2+1:mpiprocs=3
#PBS -N
JOb_batch_Q
#PBS -q
long
#pbs -j oe
#pbs -o
output_job_batch
#pbs -e
Error_job_batch
/usr/local/bin/mpirun
-np 3 --hostfile /home/jkumar/Job_test/machinefile hostname
$qsub
script.sh
|
Different Syntax Format
#PBS
-l nodes=1:ppn=2+1 (via Script)
select
1: it will choose the one node.
ncpus
2+1 : Submit 2 jobs in one node and 1 job in another node.
ncpus
3+ 2: Submit 3 jobs in one node and 2 jobs on another node.
NOTE:
Advantage is we can fire the job depends on our requirement one node
we can fire 2 job at same time another node we can fire 1 job.
Queue
Attributes.
Link
: http://dcwww.camd.dtu.dk/pbs.html
(Refer Point 6)
Syntax
:
create queue
<Queue-Name>
set queue
<Queue-Name> queue_type = Execution | Route
set queue
<Queue-Name> Priority = 40
set queue
<Queue-Name> resource_default.walltime = 72:00:00
set queue
<Queue-Name> resource_default.nodes = 1
set queue
<Queue-Name> resource_default.ncpus = 1
set queue
<Queue-Name> enabled = 1
set queue
<Queue-Name> started = 1
Note
: Resource Attributes Keywords are
walltime
,nodes ,ncpus
PBS ERROR MESSAGE
qmgr -c "set server pbs_license_info=path of license file"
Qmgr: set queue new queue_type = Execution
Qmgr: set queue new resource_default.walltime = 72:00:00
PBS ERROR MESSAGE
qmgr -c "set server pbs_license_info=path of license file"
Qmgr: set queue new queue_type = Execution
Qmgr: set queue new resource_default.walltime = 72:00:00
qmgr obj=new svr=default: Undefined attribute
qmgr: Error (15002) returned from server
Qmgr: set queue new Priority = 40
Qmgr: set queue new resource_default.walltime = 72:00:00
qmgr obj=new svr=default: Undefined attribute
qmgr: Error (15002) returned from server
SOLUTION: We have to use the Exact Keyword for resource utilization. qmgr: Error (15002) returned from server
Qmgr: set queue new Priority = 40
Qmgr: set queue new resource_default.walltime = 72:00:00
qmgr obj=new svr=default: Undefined attribute
qmgr: Error (15002) returned from server
To Resolve The ISSUE
create queue new
set queue new queue_type = Execution
set queue new enabled = True
set queue new started = True
set queue new resources_max.ncpus = 24
set queue new resources_min.ncpus = 1
set queue new resources_min.walltime=00:20:00 (20 minute)
Then the problem will be resolved.
create queue new
set queue new queue_type = Execution
set queue new enabled = True
set queue new started = True
set queue new resources_max.ncpus = 24
set queue new resources_min.ncpus = 1
set queue new resources_min.walltime=00:20:00 (20 minute)
Then the problem will be resolved.
Useful
Link:
http://www.hpc.cineca.it/content/batch-scheduler-pbs
http://www.cines.fr/spip.php?article593&lang=fr
http://hpc.sissa.it/pbs/pbs-4.html
(Related to Job Scheduler Configuration.)
PBS Unresolved Question
1)What is Check
Point In PBS.
2)What
is PBS dataservice?
3)If
the node is going down. How the job is running without any
interruption.
0 comments:
Post a Comment