The first course, Introduction to Kubernetes, is free and I found it to be a good foundation to start with.

There are video segments and follow-along style elements.

Main advice I can give is to try and get a general instanding of how the system works rather than worrying about all thr resources types.

CloudNativePG with barman backups makes it quite manageable.

deleted by creator

I personally have a K3S cluster I host at home and an auto-scaled cluster in Hetzner. I see different use cases and am happy to have both.

One thing to mention is that you can also run your own tunnel with something like pangolin on a VPS (CX23 is plenty). Thus, you could have a cheaper could bill if you wanted a hybrid setup.

Also, I highly recommend moving your node to a data center closer to home.

I’m running Immich on an old ThinkCentre with the default feature set. Currently looking at:

• Server: 1.4 GB
• ML: 288 MB
• Valkey: 8 MB
• Postgres: ~500 MB

There is a bit more overhead for the control plane and cloud-native PG operator/backup pods but my instance fits comfortably inside a ~4GB allocation.

I guess if the serious thing you’re doing is a CRUD app?

Tell that to YouTube and Spotify. Or astronauts on the ISS. Or anyone doing science really…

Machine learning and CV? Geospatial? Math in general?

God forbid they had Christmas holidays 🙄

Server is meant for all users of a single server.

So, you could buy 1-3 individual licenses or the server license for 4+ users.

deleted by creator

Basic HTML hosted at cv.dallen.co

I have a pipeline that creates a PDF version with weasyprint: https://github.com/damienallen/cv

Well, time to edit .bash_aliases…

Can’t speak to your exact machine but nowadays the license tends to be tied to the hardware.

If you are capable of manual partitioning then you should be able to reinstall Windows quickly if needed.

I would go with 16GB for the kinds of things you listed.

If you don’t need a 3.5in drive then I would go even smaller to a mini form factor. It will definitely save you money on electricity. I think it’s the main choice here.

Key-only SSH with fail2ban and I sleep easy a night.

I would go for refurb, business line SFF machines. Something like ThinkCentre or Optiplex. Specific form factor based on drive needs but the smaller you go the more power efficient. I have one on the bigger side (internal psu) that runs about 12W idle.

Just double check that it can handle hardware transcoding. Should fit right in your budget!

deleted by creator

deleted by creator

deleted by creator

deleted by creator

This level of precision doesn’t make sense. I wouldn’t go any higher than 5 digits which is already meter accuracy.

Especially in the context of a portfolio, this would count against you for geospatial software roles.