Hello all, I've hit a road block on attempts to containerize NICE-DCV alongside my NVidia drivers. Because my cluster is "strategically stuck" at RHEL6.4, I'm hoping to use a RHEL6.9 container to offer ANSYS v18.1 with 3D NVidia/Nice-enabled graphics. I was at least successful at centralizing the matching version of our K2 NVidia driver to an nfs mount point and get successful output from nvidia-smi output as follows:
Singularity rhel69_ansys182:/scratch/sandboxes_temp> nvidia-smi
Mon Feb 19 15:07:56 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 361.45.18 Driver Version: 361.45.18 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GRID K2 Off | 0000:05:00.0 Off | Off |
| N/A 29C P8 17W / 117W | 28MiB / 4095MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 GRID K2 Off | 0000:06:00.0 Off | Off |
| N/A 27C P8 17W / 117W | 98MiB / 4095MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 2 GRID K2 Off | 0000:84:00.0 Off | Off |
| N/A 30C P8 17W / 117W | 33MiB / 4095MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 3 GRID K2 Off | 0000:85:00.0 Off | Off |
| N/A 28C P8 17W / 117W | 33MiB / 4095MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
I then attempted to install NICE-DCV server inside this same RHEL6.9 container. The container is hosted on a RHEL6.4 interactive node which has a working installation of this dame NVidia driver installed, plus a working installation of NICE-DCV server. After installation of NICE in the container, I set some bind points in singularity.conf as follows with directories associated with NICE-DCV underlined below in hopes to capture my NICE license server and other files associated with NICE.:
Section from my singularity.conf file ....
# BIND PATH: [STRING]
# DEFAULT: Undefined
# Define a list of files/directories that should be made available from within
# the container. The file or directory must exist within the container on
# which to attach to. you can specify a different source and destination
# path (respectively) with a colon; otherwise source and dest are the same.
#bind path = /etc/singularity/default-nsswitch.conf:/etc/nsswitch.conf
bind path = /opt/nice
bind path = /etc/vnc
bind path = /var/lib/dcv
bind path = /usr/lib64
bind path = /etc/localtime
bind path = /etc/hosts
bind path = /scratch
bind path = /nfs/system
bind path = /nfs/prod/users
bind path = /nfs/home
When I enter my sandbox as root and attempt to enable dcv as follows, it fails with inability to find 32bit OpenGL library...
Singularity rhel69_ansys182:/scratch/sandboxes_temp> dcvadmin enable
ERROR: cannot find system 32 bit OpenGL library.
Since dcv must be enabled by root, am I chasing a false hope that NICE can work within a container for non-root container users?
Non-root user in container...
Singularity rhel69_ansys182:/scratch/sandboxes_temp> dcvadmin enable
ERROR: Only root can enable DCV.
Looking below at glxinfo inside container on the left and outside the container on the right, shows that my Nice opengl is not working properly inside the container...
Has anyone in this Singularity user group had success a getting NVidia/Nice enabled graphics to work in a container?
Thanks for any input to help. I've reached out to the vendor of NICE also, but no helpful information has been provided yet from them.
Richard