Problem writing raster predictions using raster package on Rstudio server (Supercomputer clusters)

[AbiMow]

Hi,
I am trying to generate raster predictions based on my raster stack predictors and my fitted SVM model.
I am getting the error message below:

Error: [writeStart] insufficient disk space. Estimated need: 40GB. Available: 0 GB.

I do have 2TB as primary storage. I tried to play with the rasterOptions() parameters but I still get the same error. I have also tried to change the tmpdir file location.

Thank you for your help.

[nirgrahamuk]

I'm thinking this might be less an R issue than an Operating System issue; and that File/drive Permissions will be involved.

[AbiMow]

I think I have permission to write to my primary storage space

[nirgrahamuk]

ok, lets start filling in some concrete information.
what rasterOptions did you play with ?
what tmpdir file location did you have, and then which did you try ?
(also have you run these raster predictions before on any other system and so have an understanding if 40Gb is expected/reasonable? can you run a simpler setup that would require less space?)

as for

I think I have permission to write to my primary storage space

In my experience in computing in general; often its when our basic assumptions are violated that we don't get what we expect; I've seen many occasions where a user was sure the computer was wrong, given what the user confidently knows about their data or code; which turned out to be incorrect. Is this your own personal private supercomputer? I may have assumed incorrectly that there is some sort of user access control that would be relevant but I understand how I could be wrong about that.

[AbiMow]

This supercomputer is run by my university.

You find below the parameters I was trying to modify:
chunksize" was maximized so the raster predictions can be generated in one chunk
maxmemory: to include more space for Rstudio to run my raster::predict() function
tmpdir: I have specified my own primary storage for Rstudio to store temporary files.

For your information, I already run this process on my own laptop. The size of the raster predictions was about 215Mb but the process was really slow. that's why I would like to use a high-performance supercomputer to speed up the process. I think Rstudio overestimated the required space to 40G. But, it is strange to mention in the error message "Available: 0G".

I think the error has to do with permissions to write in the primary storage space.

Sincerely yours