UT theoretical chemistry code forum

Dear Eon developers,
I am learning how to use eon with vasp. I'm interested in the example/akmc-cu100mrr-vasp. I can submit the eon job, but when it requires the vasp to run, the jobs/in directory is empty and the scripts in jobs/scratch are not submitted to the cluster. can you give advice on how to make it work? Thanks.

The akmc.log is as:

2019-09-12 14:25:37 INFO:akmc: Eon version unknown
2019-09-12 14:25:37 WARNING:statelist: State list path does not exist; Creating: .//states/
2019-09-12 14:25:37 WARNING:superbasinscheme: Superbasin path does not exist, creating .//superbasins/
2019-09-12 14:25:38 INFO:explorer: Registering results
2019-09-12 14:25:38 INFO:explorer: Processed 0 results
2019-09-12 14:25:38 DEBUG:explorer: Registered 0.0 results per second
2019-09-12 14:25:38 INFO:explorer: Queue contains 0 searches
2019-09-12 14:25:38 INFO:explorer: Making 12 process searches
2019-09-12 14:25:44 INFO:explorer: Created 12 searches
2019-09-12 14:25:44 DEBUG:explorer: Created 2.05376451319 searches per second
2019-09-12 14:25:44 DEBUG:superbasinscheme: writing
2019-09-12 14:25:44 INFO:akmc: Currently in state 0 with confidence 0.000000
2019-09-12 14:25:44 DEBUG:akmc: KMC finished in 0.000509023666382 seconds
2019-09-12 14:25:44 DEBUG:akmc: 0.00 KMC steps per second
2019-09-12 14:25:44 DEBUG:superbasinscheme: writing


Attachments are the script I used for my cluster, which uses lsf as the job manager. I submit job with the akmc-eon.lsf.

The queued_jobs.sh is:
#!/bin/sh
bjobs | awk '$1~/^[0-9]/ {print $1}'

The cancel_job.sh is:
#!/bin/sh
bkill $@

Best,
Jinfan Chen

Hi Jinfan,

First, apologies for the slow response. I have been distracted by an update of our codes from python2 -> 3, which is hopefully now almost done.

So, there are a few odd things about your input files. First, the eon program should be run directly on the master node - not through the queuing system. The eon program will queue up your eonclient jobs with the submit_job.sh script that you provide. For sge, I use a script:

#!/bin/sh
qsub -V -b y -j y -o ll_out -S /bin/bash -N "$1" -wd "$2" -pe mpi24 24 ~/code/eon/client/eonclient | awk '{print $3}'

Then, the eonclient program will run vasp directly using the runvasp.sh script. Again, I use:

#!/bin/sh
mpirun -np 24 vasp

As I mentioned for the eon server, the vasp executable should not be submitted via the queue - it is launched from the eonclient (which is run from the queue).

If you need additional help, a .tar.gz of the entire initial input files would be good. I could get it set up for our clusters as an example.

And, if it would make sense and if you are willing to give me access to the machine, I could help with the submit script files for your environment.

I am constructing a potential energy surface by calculating energy at different geometries and fitting them to an analytic function. I am using VASP package for calculating the energies. I have to generate more than 1000 points. I have different set of INCAR, POSCAR, POTCAR and KPOINTS files stored in directories like set1, set2, ... set1000. I want to use shell script to go to a directory, submit the job to the linux cluster and as soon as the job is over I want it to go the next directory and submit the next job and continue till all the jobs are done. Any suggestion on how this can be done will be very helpful.

If your linux cluster has a queueing system, you can submit a large number of jobs and let the queueing system run the jobs as the resources are free.

Another useful tool, which is implemented in most queueing systems is an array job. Here is some information about that for the sge:

https://arc.liv.ac.uk/SGE/howto/sge-array.html