1. You might want to set an appropriate bitrate and video quality for transcoding for these users that works acceptably for their latency and bandwidth.
2. The reason netflix, prime, Disney+, etc work as well as they do is because they have CDN peering, so your version of Game of Thrones comes from somewhere much closer to you in a network sense. You can change your tailscale relay, or you can lock down jellyfin to their ip and specific port and skip the tailscale (I don’t recommend it unless you know what you’re doing) and let routing tables help with latency.

I didn’t see ansible as a solution here, which I use. I run docker compose only. Each environment is backed up nightly and monitored. If a docker compose pull/up and then image clean breaks a service, I restore from a backup that works and see what went wrong.

Ah, I hear you, and sorry you had that experience. GUI controls of ZFS aren’t usually very intuitive.

Also, ZFS assumes it has direct access to the block device, and certain USB implementations (not UAS) use _sync operations that sit between the HAL and userland somewhere. So ZFS likes direct-attached storage, it’s a caveat to be sure.

If you ever change your mind, https://klarasystems.com/zfs/ has a ton of reading and tutorials on ZFS.

There’s no need to be a dick about it.

I meant no disrespect, I suppose I should have been more direct.

I asked about ZFS because it is not really that difficult to set up and there aren’t that many variables involved. You create a pool, then start using it. There isn’t much more to it.

Its not baked into the arch kernels so unless you’ve got your wits about you running updates can fuck everything up.

That is an arch problem, not ZFS. An update on Debian with ZFS would almost never behave like this.

I asked about virtualization because it would allow you to break things intentionally and flip back to a desired state, which seems to fit with your like of solving broken stuff.

So in the end, you’re obviously free to do what you like, and that’s the great thing about Linux. But you definitely seem to want to do things the hard way.

Have a better one.

2 reolink and 1 amcrest cameras, Frigate container with a coral tpu.

The reolink cameras have slightly better picture quality, but they are shit at compressing RTSP streams, so way more data flows to get the same result as the amcrest.

I didn’t downvote you, I’m genuinely curious what you mean about zfs being a “pig”.

What parts are “a bitch” to work with?

I’m a bit confuses about your approach in general:

No zfs because it “breaks”, but you use arch as server is? Sounds like you want to tinker and break things to learn, but virtualization is “overkill”?

I don’t understand what you’re trying to get from your homelab.

A pig on what?

Supply and demand.

Fair enough.

so if one dies the data is okay.

RAID1 doesn’t protect you against bitrot, config errors from 4 months ago, or a power failure that fries the whole machine.

The reason they say “RAID is not a backup” is that it is not a safeguard against the above.

I use WG tunnel, but the “turn on under X condition” doesn’t work for me.

Fair enough.

I’m speaking from both sides here, having used Plex for years and now jellyfin:

Don’t tie technical competence of a product with its monetary cost. They are not necessarily equivalent.

The problems with Plex are not technical. The problems from Plex are that they take away features and change the terms of use to the detriment of the user. Given plex’s pricing changes over the last year, I would be concerned that your lifetime pass be affected by some policy change.

I hope you’re trolling, because that is obviously not true.

The tidal range in the Bay of Fundy being 17m vs average 1m most everywhere else is due to tidal resonance from the shape and volume of the bay itself. Same with Tasiujaq.

It’s a 3rd party thing. Core functions behave well enough, ceph does not.

1. Nesting=1. This isn’t about virtualizing inside the container, it allows internal resources to access parent resources.

2. You should only need the cgroup2 entries, but they should be pointing to the correct devices:

• cgroup2 entries to allow rwm access to the correct device
• /dev/dri dir and file entries that specify bind,optional,create

Nvidia example, but quicksync is similar:

lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.mount.entry: /dev/dri dev/dri none bind,optional,create=dir
lxc.mount.entry: /dev/dri/renderD128 dev/renderD128 none bind,optional,create=file

Radicale. I just finished setting it up and with that, I can now finally shut down Nextcloud.

If the value of patient autonomy is backed up by the value of democratic decision-making, where does that leave the value of preserving life at all costs?

Strange question to ask after exploring what MAID seeks to grant.

As always, Christians fail to understand the focus of autonomy of self is the point, not your control over others’ behaviour.