I've been running two degraded ZFS arrays for the last few weeks (Debian host).

Yeah, I know. Whatevs.

One of the pools was basically "scratch" backup space and one of the spindles died (breaking the utility of the mirror). Pulled the drive, wiped the remaining, put back in service with minimum fuss as a single drive. I'll throw another spindle at it when drive prices drop again.

The other array had the SSD cache die and it's been chugging along fine ever since. Not a big deal, but from a "experience" point of view it "feels slow" like a working md array.

New SSD arrived in the mail so that'll get sorted sometime today/tomorrow.

So, what's this post about?

Linux peeps, if you are thinking about md arrays, just stop, take the time, and throw 'yer leg over the zfs horse. It's worth it.

#Linux #RunBSD #zfs #md #mdadm #raid #homelab #SelfHosted #SelfHosting

@RootMoose already have done with the no DKMS pkg as part of #alpinelinux (and one painful DKMS install on #voidlinux). So far, all good.

@EF Yeah, my main workstation (Alpine) is zfs as well (mirrored nvme).

Going forward zfs is my default for most things, perhaps excluding wee little embedded devices where the ram is small and it doesn't really matter (disposable file system). Ext4 is fine for embedded.

@RootMoose still don't get why other distros do not have a zfs package which does not rely on client side compiling.

@EF I think it's related to the uncertainty of the license, aversion to risk.

Or alternatively perhaps "not invented here syndrome".