Search code examples
parallel-processingsungridengine

cannot chdir to /path/to/job_submit_dir/ in SGE cluster

I use qsub to submit a job to the SGE cluster. In the job file, the following are defined:

    #!/bin/bash
    #
    #$ -V
    #$ -cwd
    #$ -j y
    #$ -S /bin/bash
    #

The -cwd indicates that the job will run in the directory where the job file is. All job files contains the job settings above.

Some of the jobs are submitted and could run correctly, but some of them are submitted and the status from qstat is Eqw, and when use qstat -j job_id to show the detail status, it shows:

    failed changing into working directory because: 
    error: can't chdir to /path/to/job_submit_dir

But sometimes I go into the directory, and resubmit the job, it seems to work.

I've searched in Google, and this site has provided a solution, but it doesn't work for my setting.

Could anyone give some advice, please?

Solution

  • Appears that for this instance of this error issues may be due to excessive write to network mounted storage:

    https://www.icts.uiowa.edu/confluence/display/ICTSit/Best+practices+for+high+throughput+jobs

    To solve attempt to redirect output to local storage on each execution node or /dev/null.