nix-installer v0.10.0 fails when installing Nix 2.15.0 in Alpine Linux container in Proxmox 8.0.3.

[derekmahar]

nix-installer v0.10.0 fails when installing Nix 2.15.0 in Alpine Linux container.

curl --proto '=https' --tlsv1.2 -sSf -L https://install.determinate.systems/nix/tag/v0.10.0 |
  sh -s -- install linux --init none

Error

Error: 
   0: Install failure
   1: Error executing action
   2: Action `configure_nix` errored
   3: Action `setup_default_profile` errored
   4: Failed to execute command with status 1 `HOME="/root" NIX_SSL_CERT_FILE="/nix/store/ba4f8msp39cfvfpw3m7fsalb4psw347z-nss-cacert-3.83/etc/ssl/certs/ca-bundle.crt" "/nix/store/556bgbl9z8jm7030g1b31d0cblf09036-nix-2.15.0/bin/nix-env" "-i" "/nix/store/556bgbl9z8jm7030g1b31d0cblf09036-nix-2.15.0"`, stdout: 
      stderr: installing 'nix-2.15.0'
      �[31;1merror:�[0m setting uid: �[35;1mInvalid argument�[0m
      error: cannot kill processes for uid '872415232': failed with exit code 1

Metadata

key	value
version	0.10.0
os	linux
arch	x86_64

$ uname -a
Linux alpine-1 6.2.16-3-pve #1 SMP PREEMPT_DYNAMIC PVE 6.2.16-3 (2023-06-17T05:58Z) x86_64 Linux

[Hoverbear]

Hmm this may be related to auto-allocate-uids. I'll try to reproduce it. It may require some fix upstream.

[bketelsen]

same error for LXD containers

[derekmahar]

same error for LXD containers

The Alpine Linux container in which I tried to run nix-installer is an LXD container.

[manan-gup]

Same for an ArchLinux container using Distrobox with "--init" flag.

[nexhub]

I've got the same error on a Proxmox 6.4 host with an Ubuntu 20.04 guest, installer 0.10.0.
Edit: There are similar issues with the official installer (multi user)

[Hoverbear]

Does it work if you pass --extra-conf "sandbox = false" ? I wonder if it's sandboxing...

[Hoverbear]

This may be possible to detect with heuristics.

Problem machine:

concrete-woodcock% cat /proc/self/uid_map 
         0     165536   10000001

Non-problem machine

❯ cat /proc/self/uid_map
         0          0 4294967295

We may be able to set start-id https://nixos.org/manual/nix/stable/command-ref/conf-file.html#conf-start-id to something else and also configure id-count https://nixos.org/manual/nix/stable/command-ref/conf-file.html#conf-id-count, we want id-count to be at least 65536. @edolstra suggests start-id may need to be a multiple of id-count.

So on concrete-woodcock you could set

start-id = 131072
id-count = 65536

And it would start working.

[derekmahar]

Does #638 resolve this issue?

[derekmahar]

Does #638 resolve this issue?

nix-installer 0.14.0 successfully installed Nix 2.18.1 in my Alpine Linux container.

[manan-gup]

The installation works in containers now but if I try anything after that, like using the nix run command with a package, it gives me a "No space left on device" error. Both with and without an init system. Don't know what I'm doing wrong.

[derekmahar]

The installation works in containers now but if I try anything after that, like using the nix run command with a package, it gives me a "No space left on device" error. Both with and without an init system. Don't know what I'm doing wrong.

That's strange. Are you running Nix inside a container, like Alpine Linux, that doesn't use systemd? In such containers, you must run Nix using sudo nix.

[manan-gup]

I tried in an Alpine and also an Arch container. I did use sudo when there was no init system but got the same error. The error without sudo is about permissions and not device storage.

[manan-gup]

I'm getting this error with nix-installer on my host OS as well. I also used the official nix bash installation script
and I'm getting similar storage related errors. There's more than 50 GB free on my system, so, I'm stumped here.

On running nix run nixpkgs#hello I get something like:

error:
       … while fetching the input 'github:NixOS/nixpkgs/nixpkgs-unstable'

       error: creating directory '/nix/store/tmp-3165-0/x/pkgs/tools/misc/sl': No space left on device

I guess I should ask about this at a more appropriate forum but any help is appreciated.

[derekmahar]

Is it possible that your root partition or specifically /nix is mapped to a different device or volume that is full? What does df report?

[manan-gup]

No, its on my main disk. It might be a problem with setting $TMPDIR but setting it manually isn't fixing the problem. Anyway, thanks for the help.

[Hoverbear]

If you're facing this issue with the official bash scripts as well, could I invite you to open an issue on the Nix repo? :)

[manan-gup]

Yeah sure. I'll do it tomorrow. Night time here. 🙂