I’ve got a Quasar SSR web app that is failing on deployment due to post-processing scripts that tap into Web/OpenGL properties. My current VPS is DigtialOcean, and as of this writing, they don’t have GPU support. Of the other official VPS integrations, Linode has an invite-only program to startup a GPU server. It can be done on AWS if only I wasn’t a mere mortal incapable of understanding their circular docs…
However I successfully created a GPU+ Ubuntu server using Paperspace, but I can’t get that server to integrate with Cleavr because I can’t SSH as root to execute the mkdir command. (I can SSH with a unique user that has the same privileges, but not as ‘root’).
Has anyone here come across a similar issue? Finding a VPS that offers GPU support that Cleavr can integrate with?
I can’t blame you for having your head spin with AWS docs. We have all been there.
We don’t have experience with spinning a GPU server or with Paperspace. I wonder why they restrict the root account if they have an equally powerful substitute for it.
Unfortunately, we don’t have a way to substitute root right now but I’m adding this to the backlog.
I don’t mean to be nosy but I have a local server running a Coral GPU for surveillance purposes and I use Quasar SSR for web projects so I just got curious on your usage of a GPU?
@peterc
I’m in uncharted waters with a heavy backpack of unknowns–first timer doing SSR. This is for an e-commerce site where I intend to use 3D js frameworks and custom WebGL shaders throughout the experience. Everything was fine in dev and server testing until I started using post-processing scripts. Once I did that, my dev env was fine but my server would ping 500 (with that router error Quasar inserts to catch all errors).
Now, because this server error happened when I tried to use scripts that can tap into a computer GPU, I thought maybe my issue was because I’m using a DigitalOcean drop that is CPU-only. So I went out searching and found a VPS that can provide GPU. Alas, I’m now running two servers from two different VPS on Cleavr, one CPU the other GPU, and both are returning the same 500 error.
I’m beginning to understand my issue is webpack related. I can’t import those post-processing scripts directly from the contents of the node_module dependency that I installed and expect it to work on the server. I need to either figure out how to resolve those paths in the quasar.conf.js file (my preferred method by not sure how to do at the moment), or re-package those scripts as a proper boot file that gets installed like a module (I actually think I can make this approach work but there are so many cons to separating vendor code from the actual vendor module).
That said… even though my site is a white screen at the moment, it is stunning how much faster my 1 GPU server in Iceland is compared to my CPU drop here stateside when both fetch the same repo update to build and deploy. Like, crazy stupid fast.