I made a ZFS TUI app that I think people here will find interesting. It shows host/pool overviews, performance nuts and bolts, dataset details, including snapshots and storage reclamation analysis. I’m looking for feedback on it, especially around the presentation side: what did I get wrong, what’s not obvious?
I made packages for the usual suspects (deb, rpm, etc) but since it’s Go, you can just go install it. It’s only ever been tested on Linux, should work on a Mac (for remote hosts); won’t work on native Windows since Windows OpenSSH doesn’t do ControlMaster multiplexing (WSL is fine).
Apologies for the self-promotion, I’m just really happy how it turned out and would love critical feedback. Here or over at GH.
Wow, that’s a really cool setup and seems well thought out at first glance. I definitely plan to try it out, though it’s going to be a bit before I have time.
Excellent work. And I like your choice on stats. Dirty avg, peak and delays is useful stuff, especially if you like to dump a lot of async stuff into memory to make TXGs super smooth.
I only have 2 pools left with ZFS, but I used htop with the “recently” added ZFS meters to get a quick overview on ARC, MRU, MFU and dirty levels.
This is really useful for everyone. Covers a lot we usually get with zpool iostat, arcstat, etc.
hey, you even included ghost mru/mfu and txg sync time, that very specific nerdy stats (I love)…if anything, it’s probably too much information. I personally like to see CPU utilization on the corresponding machine because I run with zstd compression. Oh no, you got CPU and temps as well, nevermind
WHAT IS THIS HERESY. Seriously what are you migrating to? I’m in the other boat, currently looking into root-on-ZFSing my raspberry PIs too. (In all fairness that will probably be netboot-off-a-zvol instead - but still, zfs.)
And yeah the idea was to have an at-glance “is the fleet OK” view (which should be evident in the hosts column as all warnings bubble up to it) but then enough info in the inspector to actually let you make decisions without having to branch out to other tools.
Temps were kind of overkill, but since I wanted to have smart polling for drive health (just the 4-5 most common indicators for early signs of trouble), temps kind of came free with it.
And yes I realize it might be a serious case of TMI especially when looking at a pool, but at least it auto-scales; columns drop off in the order of priority and I tried to keep the most relevant stuff above the less-so; you might have to scroll for the less interesting details.
Ceph, built a cluster a year ago. Mostly because I wanted to, not because there was a technical need for it, although Ceph offers nice things and fun tinkering with. homelab hobby stuff and spending too much money on too fancy stuff for my own good, the L1T way of life
But for a single server or local machine…can’t beat ZFS.
