MPI pbs输出文件在运行期间可用



我正在运行一个程序,并且使用调试模式。在调试模式下,程序会打印出一些有用的信息,我想在运行过程中检查一下。但是,我在输出目录中找不到输出文件。。它在代码中可用吗?或者我可以在代码末尾获得它。

有作业信息:

Job Id: 420302.kavun-ib
Job_Name = pKmeans
Job_Owner = e1893122@nar-ib
resources_used.cput = 00:00:00
resources_used.mem = 8148kb
resources_used.vmem = 138560kb
resources_used.walltime = 20:27:15
job_state = R
queue = cenga
server = kavun-ib
Checkpoint = u
ctime = Wed Sep  3 14:40:27 2014
Error_Path = nar:/home1/e1893122/10m_data/pKmeans.err
exec_host = ceng33-ib/7+ceng33-ib/6+ceng33-ib/5+ceng33-ib/4+ceng33-ib/3+ce
ng33-ib/2+ceng33-ib/1+ceng33-ib/0+ceng34-ib/7+ceng34-ib/6+ceng34-ib/5+
ceng34-ib/4+ceng34-ib/3+ceng34-ib/2+ceng34-ib/1+ceng34-ib/0+ceng35-ib/
7+ceng35-ib/6+ceng35-ib/5+ceng35-ib/4+ceng35-ib/3+ceng35-ib/2+ceng35-i
b/1+ceng35-ib/0+ceng36-ib/7+ceng36-ib/6+ceng36-ib/5+ceng36-ib/4+ceng36
-ib/3+ceng36-ib/2+ceng36-ib/1+ceng36-ib/0+ceng37-ib/7+ceng37-ib/6+ceng
37-ib/5+ceng37-ib/4+ceng37-ib/3+ceng37-ib/2+ceng37-ib/1+ceng37-ib/0+ce
ng38-ib/7+ceng38-ib/6+ceng38-ib/5+ceng38-ib/4+ceng38-ib/3+ceng38-ib/2+
ceng38-ib/1+ceng38-ib/0+ceng39-ib/7+ceng39-ib/6+ceng39-ib/5+ceng39-ib/
4+ceng39-ib/3+ceng39-ib/2+ceng39-ib/1+ceng39-ib/0+ceng40-ib/7+ceng40-i
b/6+ceng40-ib/5+ceng40-ib/4+ceng40-ib/3+ceng40-ib/2+ceng40-ib/1+ceng40
-ib/0
Hold_Types = n
Join_Path = n
Keep_Files = n
Mail_Points = e
Mail_Users = "xxx@xxxx"
mtime = Wed Sep  3 14:42:13 2014
Output_Path = nar:/home1/e1893122/10m_data/pKmeans.out
Priority = 0
qtime = Wed Sep  3 14:40:27 2014
Rerunable = True
Resource_List.cput = 14400:00:00
Resource_List.mem = 20gb
Resource_List.nodect = 8
Resource_List.nodes = 8:ppn=8
Resource_List.walltime = 9600:00:00
session_id = 4383
Variable_List = PBS_O_HOME=/home1/e1893122,PBS_O_LANG=en_US.UTF-8,
PBS_O_LOGNAME=e1893122,
PBS_O_PATH=/usr/mpi/gcc/openmpi-1.2.8/bin:/usr/kerberos/bin:/usr/loca
l/bin:/bin:/usr/bin,PBS_O_MAIL=/var/spool/mail/e1893122,
PBS_O_SHELL=/bin/bash,PBS_SERVER=nar,PBS_O_HOST=nar,
PBS_O_WORKDIR=/home1/e1893122/workspace/parallel-kmeans,
PBS_O_QUEUE=cenga
etime = Wed Sep  3 14:40:27 2014
submit_args = pKmeans.pbs
start_time = Wed Sep  3 14:40:57 2014
start_count = 1

如果您的输出文件总是在网络安装的位置,那么您可以要求管理员设置

$spool_as_final_name true

在妈妈的配置文件中。这将使输出直接写入其最终位置,而不是在其他地方进行后台处理,然后复制。

文件的位置取决于Torque/PBS的配置方式。有时在作业工作目录中创建,有时在计算节点上创建,并在作业完成后复制到工作目录。如果是这种情况,您应该能够在PBS/Torque spool目录中找到该输出/var/spool/torque/spool和/var/lib/ttorque/sool是两个开始查找的地方。你也可以通过将你的输出重定向到你指定的文件来解决这个问题

mpirun -np 8 ./a.out > /home/neouyghur/job_output

mpirun -np 8 ./a.out &> /home/neouyghur/job_output

获取stdout和stderr。

您可能还想查看qsub手册页中的o、e、j和k开关,因为它们也会影响输出的处理方式。

最新更新