wrf提交作业后，能显示任务，但一直不执行、没有结果

兰北

本人初学wrf，通过服务器，如果直接在xshell的前端输入./real.exe,可以正常输出边界与初始值，但是通过qsub提交作业后，虽然能生成作业编号，但是一直没有结果输出，找了好久，也不知问题出在哪里，请各位大神多多指教！谢谢！


pbs文件如下：

#!/bin/sh
#################################################################
#job name
#PBS -N WRFemreal

#PBS -d /disk4/wangll/wrf/WRFV3/test/em_real

#nodes: number of nodes requested by job
#ppn  : the number of processors per node requested by job
#eg   : nodes=4:ppn=4
#eg   : nodes=cu01:ppn=4+cu02:ppn=8
#eegg PBS -l nodes=cu03:ppn=2+cu04:ppn=2
#eegg PBS -l nodes=8:ppn=8
#PBS -l nodes=cu15:ppn=8+cu17:ppn=8+cu19:ppn=8+cu20:ppn=8

#Maximum amount of real time during which the job can be in the running state
#PBS -l walltime=48:00:00

#Maximum amount of physical memory used by the job
#PBS -l mem=8000mb

#set the queue
#PBS -q routine

#Defines the set of conditions under which the execution server will send a mail
#message about the job.
#PBS -m abe

#Declares the list of users to whom mail is sent by the execution server when it
#sends mail about the job.

#Declares that all environment variables in the qsub commands environment are
#to be exported to the batch job
#PBS -V

#executable programe
EXEC=/disk4/wangll/wrf/WRFV3/test/em_real/real.exe
#EXEC=/disk4/wangll/wrf/WRFV3/test/em_real/wrf.exe


################################################################
NP=`cat $PBS_NODEFILE | wc -l`
NN=`cat $PBS_NODEFILE | sort | uniq | tee /tmp/nodes.$$ | wc -l`
cat $PBS_NODEFILE > /tmp/nodefile.$$
#sed -i s/$/-ib/ /tmp/nodes.$$
#sed -i s/$/-ib/ /tmp/nodefile.$$

#awk '{ ++s[$1] } END {for(a in s) print a":"s[a]}' $PBS_NODEFILE > /tmp/nodefile.$$

mpdboot -n $NN -f /tmp/nodes.$$ -r ssh
mpiexec -genv I_MPI_DEVICE rdma -machinefile /tmp/nodefile.$$  -n $NP $EXEC

mpdallexit
rm -f /tmp/nodes.$$

   通过qstat查询后，显示time use一直为0，如下图

二爷的屁股

帮顶~

兰北

二爷的屁股 发表于 2014-7-7 22:22
帮顶~

谢谢！要顶起来啊

薇风拂尘

换几个节点试试 有坏的地方

兰北

薇风拂尘 发表于 2014-7-8 09:41
换几个节点试试 有坏的地方

换了节点了，还是不行，难道是队列的问题 routine需要修改？

sunshinezhpb

showbf -v 查看节点状态吧，估计计算节点全跪了

兰北

sunshinezhpb 发表于 2014-7-8 16:07
showbf -v 查看节点状态吧，估计计算节点全跪了

查看了一下，是这样

node cu01 is unavailable (state 'Busy')
node cu02 is unavailable (state 'Busy')
node cu03 is unavailable (state 'Down')
node cu04 is unavailable (state 'Down')
node cu05 is unavailable (state 'Busy')
node cu06 is unavailable (state 'Down')
node cu07 is unavailable (state 'Down')
node cu08 is unavailable (state 'Down')
node cu09 is unavailable (state 'Busy')
node cu10 is unavailable (state 'Down')
node cu11 is unavailable (state 'Down')
node cu12 is unavailable (state 'Down')
node cu13 is unavailable (state 'Down')
node cu14 is unavailable (state 'Down')
node cu15 is blocked by reservation NONE in   INFINITY
node cu16 is unavailable (state 'Down')
node cu17 is blocked by reservation NONE in   INFINITY
node cu18 is unavailable (state 'Down')
node cu19 is blocked by reservation NONE in   INFINITY
node cu20 is blocked by reservation NONE in   INFINITY

兰北

sunshinezhpb 发表于 2014-7-8 16:07
showbf -v 查看节点状态吧，估计计算节点全跪了

节点这种状态，表示什么意思呢？估计是节点的问题啊

sunshinezhpb

兰北 发表于 2014-7-9 07:55
节点这种状态，表示什么意思呢？估计是节点的问题啊

就是节点的问题了，20个就4个能用，联系管理员看看吧。