Issue connecting worker to the master

Install, Configure and Update
1

I used Command " qsub -I -q gpu -l select=1:ncpus=128:ngpus=4" to log on to a gpu node in cluster.
There I installed Cryosparc worker. Now I have facing issue connecting worker to the master.

[dpatra@r05gn02 ~]$ cd cryosparc/cryosparc_worker/
[dpatra@r05gn02 cryosparc_worker]$ ./bin/cryosparcw connect --worker r05gn02 \
>                          --master champ2 \
>                          --port 40000 \
>                          --nossd \
>                          --update \
>
 ---------------------------------------------------------------
  CRYOSPARC CONNECT --------------------------------------------
 ---------------------------------------------------------------
  Attempting to register worker r05gn02 to command champ2:40002
  Connecting as unix user dpatra
  Will register using ssh string: dpatra@r05gn02
  If this is incorrect, you should re-run this command with the flag --sshstr <ssh string>
 ---------------------------------------------------------------
  Connected to master.
 ---------------------------------------------------------------
  Current connected workers:
 ---------------------------------------------------------------
  Worker will be registered with 128 CPUs.
Traceback (most recent call last):
  File "/home/dpatra/cryosparc/cryosparc_worker/bin/connect.py", line 147, in <module>
    assert len(target) > 0, "Worker %s has not been registered so cannot be updated." % worker_hostname
AssertionError: Worker r05gn02 has not been registered so cannot be updated.
[dpatra@r05gn02 cryosparc_worker]$
[dpatra@r05gn02 cryosparc_worker]$ =>> PBS: job killed: walltime 1827 exceeded limit 1800
logout
qsub: job 88468.champ1 completed.

Also I have a question. I installed Cryosparc worker on r05gn02 node so if the node is busy can I run cryosparc?

2

The cryosparcw connect --update option should only be used when a worker with the hostname given with the --worker parameter has previously been connected/registered in the database, but …

Given that r05gn02 is part of a PBS cluster, it may be preferable to

  1. install cryosparc_worker/ on storage that is shared between cluster nodes (if /home/dpatra/cryosparc/cryosparc_worker/ is not so shared already)
  2. instead of using cryosparc_worker/bin/cryosparcw connect for a worker node connection, connect a worker cluster using
    cryosparc_master/bin/cryosparcm cluster connect (details).