Failed to launch! 255

Install, Configure and Update
1

One of my GPU desktops, running in standalone mode suddenly stopped accepting jobs after a major crash. I decided to upgrade to the most recent version of cryosparc and created a new database. The master works fine (import jobs run ok), but the worker is refusing to accept jobs. The error message is pasted below. In the process I had to create a new user, identical to the original user (which disappeared after rewriting the database).

“cryosparcm listusers” prints out a list with a single user, as expected…

As suggested elsewhere on this forum, I added the following lines to the end of cryosparc_master/config.sh:
export CRYOSPARC_MASTER_HOSTNAME=localhost
export CRYOSPARC_HOSTNAME_CHECK=localhost

Is there something I can do before reinstalling from scratch?

Peter

License is valid.

Launching job on lane default target sonja …

Running job on remote worker node hostname sonja

Failed to launch! 255
ssh_askpass: exec(/usr/libexec/openssh/ssh-askpass): No such file or directory
Permission denied, please try again.
ssh_askpass: exec(/usr/libexec/openssh/ssh-askpass): No such file or directory
Permission denied, please try again.
ssh_askpass: exec(/usr/libexec/openssh/ssh-askpass): No such file or directory
Permission denied (publickey,gssapi-keyex,gssapi-with-mic,password).

2

Very likely, if this response is not too late :ambulance: .

This should work if

  1. You don’t plan adding additional worker nodes to this CryoSPARC instance in the foreseeable future.
    CRYOSPARC_MASTER_HOSTNAME=localhost would not be compatible with a multi-node CryoSPARC instance.
  2. The worker record inside your database has "hostname": "local". To check, please can you run the command
    cryosparcm cli "get_scheduler_targets()"

As you

you may want to check for and comment out duplicate definitions of these variables higher up in cryosparc_master/config.sh, to avoid confusion in the future.

1 Like
3

Thank you, this helped!

I removed current worker (‘sonja’) and added a new worker ‘localhost’:

cryosparcm cli ‘remove_scheduler_target_cluster(“‘sonja’”)’

./bin/cryosparcw connect --worker localhost --master sonja --port 61000 --ssdpath /scratch/cherep01/cryosparc --gpus 0,1

And now it works!

THANK you for your help!

Peter

4