[molpro-user] 答复: I/O error / No space left on device

[Manhui Wang]

Hi,
Here is an example to set up more scratch directories in .bashrc:
export TMPDIR=/scratch/$USER/molpro1/$HOSTNAME

export TMPDIR2=/scratch/$USER/molpro2/$HOSTNAME
or 
export TMPDIR2=/scratch2/$USER/molpro2/$HOSTNAME
if [ ! -d ${TMPDIR} ]; then
   mkdir -p ${TMPDIR}
fi
if [ ! -d ${TMPDIR2} ]; then
   mkdir -p ${TMPDIR2}
fi

The benefit is that it ensures every job has its own two scratches and can avoid the unnecessary disturbance by other users and other Molpro jobs.
Normally Molpro can delete all temporary files belonging to the job after it ends successfully. When a job dies, Molpro also tries to delete them, but some corrupted files may still exist there.
Could you check the /scratch space at the time the job was dying? It will help you determine whether it is really a problem without enough space.

Best wishes,
Manhui

-----molpro-user-bounces at molpro.net 撰写:----- 
收件人: molpro-user at molpro.net
发件人: psc 
发件人: molpro-user-bounces at molpro.net
日期: 18/03/2011 14:03
主题: [molpro-user] I/O error / No space left on device

Hello, our computational chemists have problem with dying jobs due to "no space" on device ,but there is plenty (about 1TB) in the location where the scratch files are stored and that's where the scratch file/s are written.  

They run in to this problem only when the calculation switches to the largest basis set.  They said that "everything works up until I reset the basis to basis = aug-cc-pVTZ "
 
This is the error they get:
Write error in iow_direct_write; fd=19, l=32768, p=125397372928; write returns -1 
This may indicate a filled disk, or that a disk quota has been exceeded    (there are no quotas on this system) 
forrtl: No space left on device   (but there is about 1TB left in the scratch space)

This is tail from the output_error.txt file 
ERROR WRITING        32768 WORDS AT OFFSET************ TO FILE 4  IMPLEMENTATION=df   FILE HANDLE=  1019  IERR=-77824
 ? Error
 ? I/O error
 ? The problem occurs in writew
.. it's raid10 on four disks with no errors whatsoever


* In /etc/profile we specified only one location for scratch since the scratch space is in raid10 with fast FS
export TMPDIR=/scratch


Any ideas? Thanks for your help to tackle this problem.

Have a great day,
psc
_______________________________________________
Molpro-user mailing list
Molpro-user at molpro.net
http://www.molpro.net/mailman/listinfo/molpro-user
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20110318/ef8062e5/attachment.html>