Hi all,
the install of v2.2 on Centos 7, with cuda 9.2 seems to have gone fine, but when I launch my first job I get a “license invalid” error. Using the same license ID that I used during installation.
Please advise,
Is it possible if you can copy & paste the exact inputs+outputs that got you here? A screenshot would also work.
hi @stephan,
I can also send you the metadata associated with this job, the first step of running through the tutorial.
I have already tried cryosparcm stop, update, re-start, etc. and always get the “License is invalid” error.
We’ve confirmed your cryoSPARC v2 license is valid and should work. Is it possible that the machine running the cryoSPARC master instance doesn’t have access to the public internet to contact our servers?
- Suhail
there are definitely ucsf firewalls in place that I don’t understand my way around. But I can surf the internet and download files from the outside work, e.g. sbgrid, without problems. I can also ssh into the machine from other UCSF-networked machines. Can you advise me on where to look for the relevant config files, or a test I can perform to find out?
Update. I deleted everything and started over using the stand-alone instructions. Now the install is not behaving as expected, it crashes here:
Completed.
Currently checking hash for ctffind
Dependencies for ctffind have changed - reinstalling…
Completed.
Completed dependency check.
******* CRYOSPARC WORKER INSTALLATION COMPLETE *******************
In order to run processing jobs, you will need to connect this
worker to a cryoSPARC master.
Connecting the cryoSPARC worker to the master…
CRYOSPARC CONNECT --------------------------------------------
Attempting to register worker localhost to command localhost:39002
Connecting as unix user frosty
Will register using ssh string: frosty@localhost
If this is incorrect, you should re-run this command with the flag --sshstr
Connected to master.
Current connected workers:
Autodetecting available GPUs…
Traceback (most recent call last):
File “bin/connect.py”, line 191, in
gpu_devidxs = check_gpus()
File “bin/connect.py”, line 73, in check_gpus
num_devs = print_gpu_list()
File “bin/connect.py”, line 7, in print_gpu_list
cudrv.init()
pycuda._driver.Error: cuInit failed: unknown error
I know my RSA key is ok because this works without a password:
[frosty@localhost cryosparc2_master]$ ssh frosty@localhost
Last login: Sat Sep 29 20:20:11 2018
Is it possible if you can try installing CUDA 9.1 on your machine, and then installing the worker node again?
How about cuda 8.0? I have that through my sbgrid installation. Is there a reason to favor 9.1 over 8.0?
Nope! You can try 8.0.