Rune @rune@mcd.dk

Note to self: fediverse loves posts about feet.

Something about Maslow's hierarchy of needs where the pyramid is just "pictures of feet" all the way down

You can now text a phone number to get machine learning generated picture of a pair of feet that doesn't exist.

I think human progress has peaked

Me listening to BBC radio 2: "The British have gotten really whiny after Brexit"

And it works totally fine

But.. This IS windows 10???

Weird how some semi-old go projects are getting that haskell feel where it's supposed to "just build" but it takes me significant time and effort to get it to build.

So the only benefit to pulling ubuntu:<tag> is saving 3 lines in your dockerfile.. and trusting a 3rd party to MITM you for no good reason.

Huh, turns out the dockerhub ubuntu images are just tar files downloaded from https://partner-images.canonical.com/oci/

And 3 lines of Dockerfile.
https://github.com/tianon/docker-brew-ubuntu-core/blob/b5a2c2a86e9dc83ef3a21551ea97e351697ed2d7/impish/Dockerfile

@samgai make tons of minimal viable products and then trash nearly all of them!

@meena @rysiek the bare minimum to not get sued or regulated, of course.

@codewiz dd would probably be faster, but so much less convenient when you suddenly have no fedora installations

@dredmorbius It's "serial" from the perspective of the minio client.

The problem is that I was clever and put 16 "disks" on one pool. So every file is read from 16 different folders on the same pool.

You can only get minio's replication feature if you enable erasure code/versioning and that forces you to write to 4 "disks" with a minimum of 2 parity disks, aka 200% disk usage. Setting that to 16 "disks" gives you only ~15% parity, but also terrible performance, apparently.

@ZVCorrupted I see a download button in the mastodon UI, but if that doesn't work: https://next.magnox.me/index.php/s/Af7Kiq7F3LzLegP

@alexandra I copied 16GB files from one bucket to another on one minio instance, which triggered replicating that 16GB to 2 other minio instances one of which were on the same machine.

I think the primary problem is the number of files created due to erasure code... I did not follow the best practices of the docs..

ooops..

"01:28:05 up 3 days, 9:39, 3 users, load average: 302.62, 266.64, 154.55"