[QE-users] File write problems : suspected gpfs filesystem issues

[Ian Shuttleworth]

Dear QE community

I have qe-6.2 installed using Intel/15.0.2 and OpenMPI/2.1.1-intel15

The build completes successfully and I'm able to execute pw.x from scratch
with no problems.

However, if I save the wavefunctions to disk and subsequently try to
re-read them using restart or another QE application (e.g. projwfc.x) I get

from davcio :
i/o error in davcio

The error only appears if I am using a vdW correction, though the error
arises whether or not I use wf_collect.
If use neb.x the simulation stalls after the first iteration with a similar
error, whether or not I am using vdW corrections.

I've tried the usual strategies (judging by posts already in the forum) of
varying the number of k points against the number of nodes / cores but with
no luck. I therefore suspect this problem may be arising because of the
interaction between QE and the gpfs filesystem that I'm using. I therefore
wanted to know if anyone else has experience of this type of filesystem /
file-writing problem and if so, how they configured their installation.

The best archived reference I could find about this type of issue is a
little dated, for a different system than the one I'm using and not
entirely the same issue that I've outlined, so I wanted to know if any
other group is facing problems with file writing, particularly with gpfs:

https://www.mail-archive.com/users@lists.quantum-espresso.org/msg14495.html

With thanks

Ian Shuttleworth

(Nottingham Trent University, UK)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20180605/9ac1fdbf/attachment.html>