rentar42

ZFS is nifty and I really like it on my Homelab Server/NAS. But it is definitely a "sysadmins filesystem". I probably wouldn't suggest it to anyone just for their workstation, as the learning curve is significant (and you can lock yourself into some bad decisions).

ZFS combines the features of something like LVM (i.e. spanning multiple devices, caching, redundancy, ...) with the functions of a traditional filesystem (think ext4 or similar).

Due to that combination it can tightly integrate the two systems and not treat the "block level" as an opaque layer. For example each data block in ZFS is stored with a checksum, so data corruption can be detected. If a block is stored on multiple devices (due to a mirroring setup or raid-z) then the filesystem layer will read multiple blocks when it detects such a data corruption and re-store the "correct" version to repair the damage.

First off most filesystems (unfortunately and almost surprisingly) don't do that kind of checksum for their data: when the HDD returns rubbish they tend to not detect the corruption (unless the corruption is in their metadata in which case they often fail badly via a crash).

Second: if the duplication was handled via something like LVM it couldn't automatically repair errors in a mirror setup because LVM would have no idea which of the blocks is uncorrupted (if any).

ZFS has many other useful (and some arcane) features, but that's the most important one related to its block-layer "LVM replacement".

I was about to mention that ;-) Whenever I find myself looking for software in an area where I don't know any of the "competitors", I'll look for the awesome-* lists. At the very least they are always a good starting point.

I'm torn a bit, because architecturally/conceptually the split that LVM does is the correct way: have a generic layer that can bundle multiple block devices to look like one and let any old filesystem work on top of that. It's neat, it's clean, it's unix-y.

But then I see what ZFS (and btrfs, but I don't use that personally) do while "breaking" that neat separation and it's truly impressive. Sometimes tight integration between layers has serious advantages too and neat abstraction layers don't work quite as well.

Once I've set up SSO I'd want to use it in as many places as possible. Not having to handle additional unnecessary passwords is a benefit.

Yes! As soon as your homelab grows above a couple of services and especially if it's used by two or more people SSO becomes an absolute necessity! The tolerance of non-technical users for handling a bunch of passwords and having to enter them everywhere is understandably low.

The Home Assistant devs apparently also deal SSO as "a corporate feature that big-corp interests want to force onto us" whereas it's the exact opposite in many cases: If we want self hosted services to be a realistic alternative to the "big corpo offerings" then we have to consider convenience and security an important feature and SSO is one of the few things that improves both at the same time.

This is also by the author of Uptime Kuma, which explains the similarity in UI.

Portainer has so many tiny broken places that I effectively treat it only as a read-only view. It lists my containers and shows my logs and nothing much else.

It could in theory do quite a bit more, but starting from the fact that it doesn't quite do docker-compose, but its own thing that's somehow similar but different there's just too many tiny issues with it.

Also, it's quite aggressively pushing the paid option without a way to turn that off (or at least turn it down to tolerable levels).

I'm using it too and even the current prices are reasonable (especially if you consider there's no other fees, no transfer, no ingress, no egress, ...). If you put it in S3 glacier and you ever have to restore a relevant chunk of your data (or god forbid, want to do periodic testing of the backed up data) then you'll be paying quite a bit of fees.

It doesn't exist.

Wait. Again?

I just use Ansible to prepare the OS, set up a dedicated user, install/setup Rootless Docker and then Sync all the docker compose files from the same repo to the appropriate server and launch/update as necessary. I also use it to centrally administer any cron jobs like for backup.

Basically if I didn't forget anything (which is always possible) I should be able to pick a brand new RPi with an SSD and replace one of mine with a single command.

It also allows me to keep my entire setup "documented" and configured in a single git repository.