13 comments

[ 0.15 ms ] story [ 12.2 ms ] thread
Thanks for this. So weird though - I literally spent the better part of the weekend reading and learning all about the differences between zram and zswap and ultimately deciding to switch out my previous zram config for zswap on my NixOS system! So seeing this pop up on my hackernews feed felt a bit spooky…
same, I literally just set up my Framework 13 Pro today and was looking for the current recommended zswap settings... looks like a bunch of us are in this thread
I agree with everything else here, but swap files on btrfs should probably be avoided whenever possible [0].

[0]: https://btrfs.readthedocs.io/en/latest/Swapfile.html

Right, you are good to do swap files on btrfs since linux 5.0 or so as long as you follow some rules. Disko thankfully covers all those details for you, but I shouldn't have skipped over that in the piece. I'll make an update when I get a chance.
Swap files work, but I wouldn't quite say "you are good to do swap files on btrfs" because the official documentation (the page I linked earlier) says

> filesystem - must be only single device

> This means swapfiles will have an amplified impact, much larger than the size of the swapfile itself.

> Users should be aware of the impact when using swapfiles alongside other data. This is especially discouraged for root filesystem.

If a swap file is the only thing that will work with your setup, then it's fine to use one, but if you're at all able to use a swap partition, then that's a far better choice. (Swap files are fine on ext4 and xfs; it's only COW filesystems like btrfs and zfs where they should be avoided.)

While everything else is pretty obvious from the nature of an unmovable file, "scrub will skip those block groups which have extents of an active swapfile on them, and those block groups may also contain extents belonging to other files".

Then again, I mostly use XFS, which has no data checksums at all, so YMMV.

I've had good luck using zswap+zstd on Ubuntu 26.04. I have a Framework 13 (AMD) with 16GB, and have a 16GB swap file configured (ext4 root partition).

I had planned to upgrade the machine to 32GB of RAM but I don't really feel it's necessary at this point. Which is obviously great given the current RAM costs!

I also just recently enabled zswap on my NixOS setup, it's been working fine. I have 32GB of RAM and 64GB of swap, so I really don't need it, but it hasn't hurt.

I was expecting some problems with my switch to NixOS back in June, but it's been just superb. Even things like building deb packages for my Ubuntu 26.04 machines at work have not been a problem at all.

I had dabbled with NixOS for a year or two before making the switch and had some familiarity with the nixlang and various concepts (SOPS, age, flakes). But, what really smoothed out the switch was using Claude Code to develop the NixOS configs. Everything is easy to do that way, and any issues are easy to back out.

Claude code/Git/NixOS is such a nice system to work with.

You can build your entire system from a worktree and merge if it works or boot the old generation if it doesn't. This is great for trialing new kernels, NixOS releases, or alternate Window managers, and rolling back is trivial.

One downside is that there is only one stable NixOS at a time, and no LTS. However in practice I haven't found this to be an issue across multiple desktops, servers and SBCs.

Agreed. One additional thing I do is I have the same setup on my main work workstation and also my old work laptop, so when I do a big swath of package updates (about once a month) I do it on my "scratch monkey" first and make sure nothing obvious breaks. I started doing this after a GNOME update broke my GDM (as you say, until I booted the old generation, but still was a work interruption).

I'm using NixOS "latest" on my workstation, for no particularly good reason.

One success I had was with building up a custom vim setup. I had previously been using Lunar and Astro as an opionated vim distro with LSP+Treesitter and some other "table stakes" things a modern editor should have. But there wasn't a clear way to install Astro in NixOS when I tried it manually. So I just told Claude Code what plugins and settings I wanted in neovim, and now I have a single file that fully describes my vim setup. Before with Astro and Lunar the configs were spread out all over a bunch of files.

I was also able to get Claude Code to bundle up gitbutler-cli, which I wanted to try but so far have not. It builds it from source because there was no existing derivation. It takes a stunningly long time to build, but otherwise it has worked very well in my simple testing of it. Plus, I can build on my powerful laptop and then use that to seed my old, slow ThinkPad T470s and even slower chromebook that I also put my config on.

Not on nixOS, but I thought I would weigh in on my preferred value for swappiness. 0. For some reason when I have a month of uptime (and I suspend once a day), doing anything with the desktop environment for the first time in a few hours causes a micro stutter for maybe 100ms, as if having to swap out from disk. Turning swap off and on again (meaning no pages are swapped to disk anymore) immediately fixes it. Opening a new tab, pressing the super key even performing certain actions in a video game cause the stutter. I know the internet tells me that 0 isn't a good value, but my computer becomes noticeably worse to use even with a swappiness of 5. I have enough RAM for anything I want to use this computer for, so swap is just for suspend for me.
whether to use zram or zswap depends on what your intentions are, hibernation doesn't matter (use priority)

if your intention is to have enough compressed ram such that you never reach the limit (compression ratios are usually around 3 with zstd) you can allocate over 100% of your ram in zram. then have 100% of your ram as a swap partition for hibernation. this works extremely well but should you exhaust your zram swap serious problems may begin.

zswap will be using your SSD and putting wear on it and also imposing latency spikes on old page access, only makes sense if you are certain your swap usage may spike too high in my opinion.

> I mostly want to note that swap files can complicate hibernate.

Only in Linux - Windows has no problem with swap in a file. Swap partition is an awful idea because you cannot easily resize it.

Zram is also a bad idea, when it is filled, the system starts swapping to a disk, and old pages nobody needs are kept in RAM forever. Zswap is better in this aspect - old pages are pushed to the disk and newer ones kept in memory. But Linux distributions use zram by default.

Also, it seems that many distributions now place /tmp in RAM (tmpfs). Whoever made such decision is not very bright. When you place a large file in /tmp (for example, viewing a large compressed archive which is automatically unpacked into a temporary location), it fills the memory and the systems locks up completely, including mouse and GUI - because bright people developing Linux cannot think of reserving resources for desktop environment to prevent exactly this situation. Why reserve memory for GUI? Obviously files placed into /tmp are more important. Who needs GUI nowadays when you can easily connect over SSH from another machine and kill the misbehaving program.

If you are making a Linux distribution, do not place /tmp into memory. Or at least check how much RAM user has, and add a limit on tmpfs after which files are pushed onto the disk. Also, do reserve memory for window manager with high priority.