One downside with #NixOS that I'm noticing is that it consumes an absurd amount of hard disk space. This is especially true if you don't know to clear out old generations, but even a system with only a couple generations takes more space than I would expect: I have a VM running right now with 3 generations and a /nix folder that's 18G. That's AFTER optimizing with hard links and garbage collecting. #nix
@scott Try using btrfs and running a tool to do deduplication.
@scott yes, I confirm. I use something like this
services.beesd.filesystems = {
root = {
spec = "LABEL=nixos";
hashTableSizeMB = 512;
verbosity = "crit";
extraOptions = [ "--loadavg-target" "2.0" ];
};
};
and it saves a lot of space, without affecting CPU and RAM performances because it is a fast, incremental, off-line deduplication.
@yisraeldov
I used it on a zstd compressed filesystem and the results were above my expectations.
After installation, it had to read the entire disk, and it is slow. But the incremental update phase is unnoticeable.
@yisraeldov
I don't know, but deduplication is based on the recognition of similar blocks of data. So a defrag should not affect the deduplication, because you are defraging blocks, not "files".
I never used defrag on BTRFS, but I used a lot this:
let # BTRFS maintanance.
# Add to ``systemd..services`` settings.
balanceBtrfs = fs: when: {
description = "Fast and incremental balance of btrfs file-system ${fs}";
startAt = when;
serviceConfig = {
Type = "oneshot";
ExecStart = "${pkgs.btrfs-progs}/bin/btrfs balance start -dusage=50 -dlimit=2 -musage=50 -mlimit=4 ${fs}";
};
};
...
in ...
# Btrfs maintanance
# Without this, the free space is not reclaimed after many deletions.
systemd.services.balanceBtrfsRoot = (balanceBtrfs "/" "daily");
systemd.services.balanceBtrfsHDD1 = (balanceBtrfs "/mnt/hdd-int-1/" "daily");
services.fstrim.enable = true;
services.fstrim.interval = "weekly";
@mzan @scott Cool! I didn't know about this, I was doing dedup manually. Any idea how this works with compressed filesystems?