VASP executable not found when running Si test workflow

[Nicholas Wagner]

Hello,

I am trying to perform the Si structure optimization test workflow on the Stampede2 cluster, but I am getting a traceback that seems to indicate the vasp executable I thought I specified is not being used. On Stampede2, we have to specify "ibrun vasp_std", 

but the custodian seems to have ended up looking for just "vasp". This is despite specifying my vasp_cmd in the my_fworker.yaml file. I have copied the excerpt from the launches collection on my fireworks database below. Is there somewhere else I need to 

force atomate to use the vasp_cmd I want? Like when I create the workflow?

Best,

Nick Wagner

>>>>>>>>

{

    "_id": {

        "$oid": "598fd760915f24b1bc05eb95"

    },

    "fworker": {

        "name": "stampede2",

        "category": "",

        "query": "{}",

        "env": {

            "db_file": "/home1/03229/tg825281/anaconda3/envs/atomate_env/config/db.json",

            "vasp_cmd": "ibrun vasp_std -n 128"

        }

    },

    "fw_id": 3,

    "launch_dir": "/scratch/03229/tg825281/block_2017-08-13-04-05-29-301218/launcher_2017-08-13-04-36-49-070734",

    "host": "c428-064.stampede2.tacc.utexas.edu",

    "ip": "206.76.200.16",

    "trackers": [],

    "action": {

        "stored_data": {

            "_message": "runtime error during task",

            "_task": {

                "vasp_cmd": "vasp",

                "job_type": "double_relaxation_run",

                "max_force_threshold": 0.25,

                "ediffg": null,

                "auto_npar": ">>auto_npar<<",

                "_fw_name": "{{atomate.vasp.firetasks.run_calc.RunVaspCustodian}}"

            },

            "_exception": {

                "_stacktrace": "Traceback (most recent call last):\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/codes/fireworks/fireworks/core/rocket.py\", line 253, in run\n    m_action = t.run_task(my_spec)\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/codes/atomate/atomate/vasp/firetasks/run_calc.py\", line 185, in run_task\n    c.run()\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/codes/custodian/custodian/custodian.py\", line 320, in run\n    self._run_job(job_n, job)\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/codes/custodian/custodian/custodian.py\", line 376, in _run_job\n    p = job.run()\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/codes/custodian/custodian/vasp/jobs.py\", line 205, in run\n    p = subprocess.Popen(cmd, stdout=f_std, stderr=f_err)\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/lib/python3.6/subprocess.py\", line 707, in __init__\n    restore_signals, start_new_session)\n  File \"/home1/03229/tg825281/anaconda3/envs/atomate_env/lib/python3.6/subprocess.py\", line 1333, in _execute_child\n    raise child_exception_type(errno_num, err_msg)\nFileNotFoundError: [Errno 2] No such file or directory: 'vasp'\n",

                "_details": null

            }

[Anubhav Jain]

Hi Nick

From your output, your my_fworker.yaml looks to be set up correctly (the "env" key shows a key called "vasp_cmd" with your desired command). However, the actual workflow isn't using the env value for the VASP command. If you look at this part: "vasp_cmd": "vasp" - that means it is going to just directly execute "vasp". If it instead said "vasp_cmd": ">>vasp_cmd<<", that would be an instruction to take the value from env.vasp_cmd key. 

I am not sure how you created the workflow in the first place; the default workflows use ">>vasp_cmd<<" as the default. Can you point me to which Si test workflow you are referring to? (where in the code or documentation)

Best,

Anubhav

--
You received this message because you are subscribed to the Google Groups "atomate" group.
To unsubscribe from this group and stop receiving emails from it, send an email to atomate+unsubscribe@googlegroups.com.
To post to this group, send email to ato...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/atomate/45e88a9f-e8cf-44c9-b538-5ad74347d213%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Best,
Anubhav

[Nicholas Wagner]

Thanks for your reply. I am following the documentation at https://hackingmaterials.github.io/atomate/installation.html

The test workflow command there is: atwf add -l vasp -s optimize_only.yaml -m mp-149

To unsubscribe from this group and stop receiving emails from it, send an email to atomate+u...@googlegroups.com.

To post to this group, send email to ato...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/atomate/45e88a9f-e8cf-44c9-b538-5ad74347d213%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Best,
Anubhav

[Anubhav Jain]

Hi Nick

Ok understood. I actually never use "atwf" or directly load workflows from files (I and others call the Python function instead), so I didn't catch the fact that this method doesn't really make it easy to set your VASP command to ">>vasp_cmd<<" and just defaults to "vasp". I will forward this to the relevant people to see if they can design something to help.

In the meantime, you can replace the command in the documentation with:

atwf add -l vasp -p wf_bandstructure -m mp-149

This will add a workflow using the Python function atomate.vasp.workflows.presets.core.wf_bandstructure. You can also add other workflows in that same module, e.g., to run bulk modulus instead run:

atwf add -l vasp -p wf_bulk_modulus -m mp-149

Hope this gets you up and running without issues. As mentioned, we'll try to get an official "fix" to the docs as soon as possible

[Anubhav Jain]

Btw you can follow the progress of the fix here:

https://github.com/hackingmaterials/atomate/issues/143#issuecomment-322053601

[Nicholas Wagner]

Thanks for the help. I ran the bandstructure command you mentioned, and it crashed due to having the scratch_dir unspecified in the my_fworker.yaml file. Once I added the environment variable there, the calculation succeeded. Now on to the next part of the tutorial!

[Anubhav Jain]

Ok good to hear!

FYI, I updated the installation instructions to set "null" for the scratch_dir (which won't use a scratch_dir)

https://hackingmaterials.github.io/atomate/installation.html