What would cause this error?

Starting websocket server…
Log file is vnc.log
Successfully started VNC server on queue1-dy-compute-resource-1-1:5901…
Script starting…
Starting websocket server…
/var/spool/slurmd/job00012/slurm_script: line 185: /home/username/ondemand/data/sys/dashboard/batch_connect/sys/bc_desktop/dantest/output/6e79d40a-c219-40df-8282-3d2ece120fce/script.sh: No such file or directory

Hi and thanks for the post.

In order for us to best help, we need more context here. Please try to explain up front when you post what it is you are trying to accomplish and the associated config files if you have any.

From this, all I know is you have a vnc error because a script is missing (as the error says) but I’m not sure what it is you are doing. Is this for an interactive app of some kind? If so, could you please add the form.yml and the submit.yml to start?

My apologies. I’m attempting to start an interactive desktop.

sh-4.4$ cat form.yml

label: “Desktop Environment”
widget: select
- “gnome”
- “kde”
- “mate”
- “xfce”
bc_vnc_idle: 0
required: true
node_type: null


  • bc_vnc_idle
  • desktop
  • bc_account
  • bc_num_hours
  • bc_num_slots
  • node_type
  • bc_queue
  • bc_vnc_resolution
  • bc_email_on_started
    sh-4.4$ cat submit.yml.erb

template: vnc
cluster: “example2”

  • bc_account
  • bc_queue
  • bc_num_hours
  • bc_num_slots
  • bc_email_on_started

Could you do me a favor and format the yml as it is very difficult to troubleshoot in plain text that then gets marked up.

You can accomplish the raw formatting by using three backticks, then put all the code in the block and close it out with three more backticks.

You have an NFS latency issue - or no NFS $HOME storage at all. Here’s what’s happening

  • you submit the job in the web UI
  • we (OOD) writes these files (on the web node) to your $HOME directory which should be an NFS storage that’s attached to your compute nodes as well.
  • we (OOD) submit the job from the web host.
  • the job kicks off and attempts to read these files

Now at this point when the job kicks off - the $HOME drive that we wrote the files to needs to exist on the compute node as well. It either doesn’t exist as a network storage device (i.e., it’s local to that machine) at all or there’s some latency in the time we wrote the files on the web node and the time it takes for them to become available on the compute node in which the job is running.