Last week got my new epyc server with GPU running ollama and all the trimmings.

This week linked my 2 home bases with wire guard, all the subnets mesh and the wifi isolation is solid. Performance is surprisingly good considering they’re 9 time zones apart on different hemispheres.

Migrating plex to jellyfin to get hw accel working.

Also trying to get my second base multiple statics and 10gb if possible, rural fiber in Europe is unbelievably aweome, hope to drop Comcast business back home if it works.

Got someone to work with on a new company, so that’s part of this, though my day job relies on this too.

Doing that switch soon.

Plex doesn’t do hw accel well, which kind of defeats the purpose.

Same, except the irc, I have a python thing to interface.

Stealing your idea, that sounds awesome.

That’s it.

Sounds like a Chinese geek tried to make something useful, did a lot of dirty hacks to get it going.

And couldn’t properly explain because his social skills and English weren’t great.

The blobs weren’t super suspicious, just some gpld tools, basically busybox kind of stuff.

The real problem is what he made was so fucking insanely useful and needed by everyone that the standards for software skyrocketed.

Like you make a cure for cancer and everyone starts screaming at you because one of the side effects is temporary impotence.

Prison pocket.

It’s interesting, however, if you mkfs.ext4 without -E ssd, or through some weirdness in your driver chain the filsystem doesn’t know it can discard, then everything everywhere sucks for everyone, a cow fs is worse because no blocks are ever overwritten till the end, and the block map becomes a disaster while performance goes down the drain.

Nowadays this rarely happens outside of very broken USB mass storage chips.

This is why we used to have the fstrim command.

The fs does cow then releases the old block if appropriate.

The ssd has a tracking map for all blocks, it’s cow relies on a block being overwritten to free the old block.

Basically it works out the same either way.

4 is bad because parity is on one drive so no matter what happens that drive is the write bottleneck. Raid5 is basically raid4 + raid0.

5 is just fine but low safety, I run 6 always and it has basically never let me down.

Meh, ssds are basically cow by nature anyway, you have to erase large blocks, you can’t just rewrite into them.

That’s not something they natively support, it’s hard links or a copy.