Wednesday, 6 June, 2012 8:09 AM
HI,I have a windows HPC system which we would like to do a parallel rendering. I download the latest nvida drivers for the compute node(they have the same graphics card). I would like to install the drivers all compute nodes using clusrun command from the headnode. i started like this:
clusrun /all /exclude:headnode /workdir:the_shared_dir nvidia_driver.exe
From the compute node task manager, i can see the process is shown there but not take any CPU. I guess there is something wrong as the nvidia driver installation process is an interactive one.
The question is: 1. Should i provide some parameter for the exe file? If yes, what are these parameters?
2. If i was totally wrong, any command i can use to update the driver without manually do it on each compute node?
By the way, i test the clusrun with a simple bat file which will produce a dummy text file. The PS results saying they are successfully executed. But why i do not see the text file generated? Thanks a lot.
Wednesday, 6 June, 2012 5:21 PM
1. I guess you should provide some params to make the installation automatic, you have to check with nvidia about the parameters, hpc clusrun just runs whatever you give it, as if you are running that command on the compute node yourself
2. Have you specify /workdir for the clusrun? If not your text file should be sitting on the local drive of your compute nodes.
- Marked As Answer by Don PatteeModerator Friday, 10 August, 2012 11:56 PM
Thursday, 7 June, 2012 2:06 AM
Thanks for your answer, Michael.
Another question is that: whether am i able to burn the current well configured Head Node into a DVD so that i can just use the DVD if i would like to install another cluster system on somewhere else, without doing the configuration process again.
Thanks & Regards,
Monday, 18 June, 2012 4:08 PM
In theory you should, just do some search online please.