Dynamic, distributed, unreliable node cluster?

[Spor7biker]

I'm trying to come up with a way of using spot instances (or equivalent)  on vast.ai and the traditional cloud providers.

I have found some front ends, and a few dated write ups for automatically deploying, but nothing seems to cover shutting down instances when done, or managing nodes that may abruptly be removed.

I'd like to keep the front end running someplace 24/7, if can be done on the cheap, but I read the front end needs fast mysql access.
 
Starting instances manually isn't a big deal, having them shutdown and stop the running cost is.

I'm working and part time living out of a cargo van with limited solar power so running a desktop with GPUs isn't practical. I do have a few older laptops, but the hash rate leaves a lot to be desired...

Has someone figured this out yet? Or is it something to cobble together from existing guides and make a new guide for?

I'm trying to do this all on a tight budget.

Thanks

SB

[winxp5421]

Yes, this can be done with the hashtopolis project. https://github.com/s3inlc/hashtopolis

Create an image that will autostart agents that point to your persistent hashtopolis server.

Create tasks and wait for cheap compute to do its work.

It won't be one click and done kinda thing to set up but, it's entirely possible.