[Nek5000-users] running NEK5000 on a cluster with fast/slow storage subsystems. Relative paths instead of absolute

nek5000-users at lists.mcs.anl.gov nek5000-users at lists.mcs.anl.gov
Sun Jun 29 09:02:10 CDT 2014


Van,

It looks to me like you should move your files to ...vanvan/_scratch/stw
and launch your job from there.

I'm assuming you're using some type of script to launch the job, like one
of those provided in trunk/tools/scripts.

If you need, you can have the script copy SESSION.NAME to your top
directory, if that's where the compute node is looking for it.

In any case, a slight modification to the launch script is exactly what is
needed.   SESSION.NAME, which is created by the launch script
(e.g., nekbmpi blah 4, would create a SESSION.NAME file), has the
session name on the first line and the path where files are to be found
on the second.

Just modify your own nekbmpi script to have it reflect where the compute
node is to find the files.   Note that you'll all have to ensure that SESSION.NAME
is in a directory where the compute node can read it.

Paul

________________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov [nek5000-users-bounces at lists.mcs.anl.gov] on behalf of nek5000-users at lists.mcs.anl.gov [nek5000-users at lists.mcs.anl.gov]
Sent: Sunday, June 29, 2014 4:16 AM
To: nek5000-users at lists.mcs.anl.gov
Subject: [Nek5000-users] running NEK5000 on a cluster with fast/slow storage    subsystems. Relative paths instead of absolute

I am trying to use nek5000 on the cluster Lomonosov.
What is special in the configuration is that it has two levels of storage.
The basic storage of a user (i.e. his home directory) is not seen by computational nodes.
Computational nodes have their own file system which can be accessed by user through "~/_scratch" subdirectory of the user's home directory.
So "~/_scratch" is the home directory for programs running on computational nodes.

This is why I can't run nek5000 on this very cluster.

I have the following message in the log file:
...
 read .rea file
forrtl: No such file or directory
forrtl: severe (29): file not found, unit 9, file /mnt/data/users/dm4/vol12/vanvan/_scratch/stw/stw32x32.rea
...

You can see that the program tries to look for an absolute path of the file: it expanded home directory to "/mnt/data/users/dm4/vol12/vanvan" and tries to find the file in the "_scratch" subdirectory which does not exist on computational nodes.
For the program on computational nodes the path is simply "~/stw/stw32x32.rea" or "/home/users/vanvan/stw/stw32x32.rea"

Can you tell me what I have to change for a relative path instead of absolute or give the absolute path manually.

Suppose this situation can happen on other clusters.

Thank you, Van
_______________________________________________
Nek5000-users mailing list
Nek5000-users at lists.mcs.anl.gov
https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users


More information about the Nek5000-users mailing list