Thanks for the detailed reply. I found a QSFP+ DAC that says it supports IB and Ethernet.
I don’t have enough computers to set up a fabric, only the 2 I would be direct attaching have PCIE slots.

I’ve never used infiniband before so my reason for wanting to try it is just to learn what it is, and how it works. That said, some of those use-cases look very interesting, especially transporting NVMe namespaces, I didn’t know that was possible.

I was testing a custom initramfs that would load a full root into a ramdisk, and when I was going to shut down I tried to run rm -rf --no-preserve-root / to see what would happen, since I was on a ramdisk anyway. The computer would not boot after that because it nuked the UEFI options.

update: I managed to get it working, look at the edit

It recognised the disks in an ASR array, but the type is “unknown” and it fails to assemble with “Undefined RAID type (null)[1] on asr_”. So I don’t think that worked sadly.
EDIT: The RAID card I had supported RAID 5 and dmraid doesn’t, that’s probably why it’s not working.

I could not find a --discover parameter, but I tried --assemble --scan and it couldn’t find a super block.
It feels a bit frustrating to have all the data here but no way to access it, maybe a tool will pop up at some point if I hoard the disk images.
Thanks for the suggestion though.

I like it better than gitlab, gitlab is too cluttered and has loads of features I don’t need. forgejo will be a lot better when they get federation going though

I like to have different naming schemes for different device classes.
Desktop computers: Greek gods
Laptops: Elements of the periodic table
Cloud servers: Norse gods
Home servers: Planets of the solar system
Raspberry Pis: Greek titans

I was running lemmy on it too until a few days ago. I had an SSD for the database though.
oh and the gitlab instance was the straw that broke the camel’s back for the Pi, I ended up going with forgejo instead.

RPI: Actually dying

Me: Gitlab time

The bot uses smmry.com to make its summaries, it would be easy to make a fork that lowered the amount of sentences in the summary, but it might lose some important details.

EDIT: oops, completely wrong bot, I was looking at another lemmy TL;DR bot on github

EDIT2: should still be quite simple to fork the current one, but I couldn’t find where to reduce the number of sentences

I know, look at the usernames of the replies

Why would you need more than one?