Problem in linking machine

Alloy Theoretic Automated Toolkit (ATAT)
1

Dear colleagues,

I have encountered one problem in linking remote machine.
Everything goes fine except the job query.
The cluster I am using is using slurm environment. So I changed the "llq" command in the ~/.machines.rc file. But terminal prompts "sh: squeue: command not found" when I type "chl".
Of course, the "llq" does not work also.
I appreciate if someone can help!

at the end of .machines.rc file
#example for queueing system
#llq -u | wc -l | awk ‘{print (-4)/2}’ + qit
squeue -u myaccount | wc -l | awk ‘{print (-4)/2}’ + qit

best,
oliver

2

Are you running pollmach on the machine running slurm?
If not, add ssh in front of squeue:

ssh you@yourmachine squeue -u you

Does squeue work when you type on the command line? (Sorry I have to ask that (!))
what about in a sh script?

3

Thanks! Axel.
The problem solved when adding node command ahead.
The .machines.rc looks like this:
set waitbetweenpoll=60
echo 0.5 + none
node -r -s myaccount@cluster uptime | getvalue average + node -r -s myaccount@cluster
node -r -s myaccount@cluster squeue -u myaccount | wc -l | awk ‘{print (-4)/2}’ + qit

and the .ezvasp.rc
#!/bin/csh
#enter name of vasp executable here

set VASPCMD="vasp-half >> vasp.log 2>&1"

set VASPCMD="node -r -s myaccount@cluster sbatch /tmp/myaccount/run-vasp "
#enter the directories containing the pseudopotentials here
set POTLDA="/Users/myaccount/Documents/Programming/atat/glue/vasp/potential/LDA"
set POTGGA="/Users/myaccount/Documents/Programming/atat/glue/vasp/potential/GGA"
set POTPAWLDA="/software/apps/vasp/Potentials/5.2/PAW/LDA"
set POTPAWGGA="/software/apps/vasp/Potentials/5.2/PAW/PBE"
set POTPAWPBE="/software/apps/vasp/Potentials/5.2/PAW/PBE"

runstruct_vasp does copy the folder into cluster and submit the jobs, but it does not copy back anything. I haven’t figure out the problem. Do you have any suggestion?

Thanks in advance!

4

The are a couple of issues:

  1. the node command will attempt to copy files back after the batch is submitted, not after the commands in the batch file are completed (so there is nothing to copy back at that time). You would need a command that waits for the batch to complete.
  2. You need only one of the two "node…" lines in your .machines.rc file.
  3. you don’t need a "node" command in both .machines.rc and .ezvasp.rc
  4. IMPORTANT: if you can compile and run atat on the machine with the queueing system, it would be a lot easier. You would just run maps with the batch script in the background (with a "&" after it) and pollmach in the foreground (without a .machine.rc file).
5

Thanks Axel!
Now I have fixed the installation!

6

Thanks for the support Prof. Axel.
I need your help to figure out the final part of point 4 above.

My job script v533.sh contains the following:

#!/bin/bash
#SBATCH -N 1
#SBATCH -n 24
#SBATCH -J DFT
#SBATCH --error=job.%J
#SBATCH -p queue1
#SBATCH --output=job.%J

echo "using the following mpi which mpirun"
mpirun -np $SLURM_NPROCS /home/msg/raviky/wscript/vasp &> vtd.out

For independent vasp runs, I submit the job with the following syntax:

$sbatch -n 24 -p cmd.q /home/msg/raviky/atat/v533.sh

Your following statement is not clear to me:
You would just run maps with the batch script in the background (with a "&" after it) and pollmach in the foreground (without a .machine.rc file)

Do I still need to have the following in the .ezvasp.rc file?

set VASPCMD="sbatch -n 24 -p cmd.q /home/msg/raviky/atat/v533.sh"

If yes, what is the sequence and option with which maps -d& and pollmach runstruct_vasp are executed at the command prompt?

If no, what set VASPCMD= should have?
What is my modified v533.sh?
What is the typical command sequence?

What is running maps with the batch script in the background and pollmach in the foreground.
(No .machine.rc file created)