toraritte.github.io

[Ubuntu, NixOS] Resize the /run directory (and temporarily increase swap size)

A Nix custom image build kept failing with the message No space left on device on Ubuntu 18.04.4 LTS, and was able to track it down using watch -n -1 "df -h" while running the build. The culprit ended up being a /run mountpoint (more specifically, /run/user/1000) and, indirectly, the swap space.

This is how it looked like before:

Filesystem                   Size  Used Avail Use% Mounted on
...
tmpfs                        785M   28K  785M   1% /run/user/1000
...

On NixOS, instead of the steps below, the process would have been simply

  1. to edit the services.logind.extraConfig attribute in /etc/nixos/configuration.nix:
  services.logind.extraConfig = ''
    RuntimeDirectorySize=12G
    HandleLidSwitchDocked=ignore
  '';
  1. Rebuild configuration (e.g., with sudo nixos-rebuild switch).

More info:

1. What is /run? (Baby don’t hurt me)

This is the authoriative answer, but this Quora answer sums it up:

/run is the “early bird” equivalent to /var/run, in that it’s meant for system daemons that start very early on (e.g. systemd and udev) to store temporary runtime files like PID files and communication socket endpoints, while /var/run would be used by late-starting daemons (e.g. sshd and Apache).

Traditional /var/run was an actual directory on disk, which meant the underlying filesystem may not have been mounted at the point systemd et al needed to write stuff into it. Making /run a tmpfs (i.e. RAM-based) filesystem neatly solved this problem and eliminated the need to clean it up on the next boot.

Of course, having two runtime scratch directories struck many as being a bit much, so in many modern Linux distros, /var/run is just a symlink to /run.

2. tmpfs

kernel.org/doc/Documentation/filesystems/tmpfs.txt:

Everything in tmpfs is temporary in the sense that no files will be created on your hard drive. If you unmount a tmpfs instance, everything stored therein is lost.

tmpfs puts everything into the kernel internal caches and grows and shrinks to accommodate the files it contains and is able to swap unneeded pages out to swap space. It has maximum size limits which can be adjusted on the fly via mount -o remount ...

Also, “tmpfs lives completely in the page cache and on swap”.

(TODO: Read kernel docs more often. Heart of Darkness can wait.)

3. Resize /run mountpoints

According to the tmpfs documentation , “tmpfs has three mount options for sizing” where size is

The limit of allocated bytes for this tmpfs instance. The default is half of your physical RAM without swap. If you oversize your tmpfs instances the machine will deadlock since the OOM handler will not be able to free that memory.

That is to say, it can be set to an arbitrarily large size, but make sure that there is enough RAM or swap space (or combination thereof), so adjust the latter if needed (see 4. below).

In my case, I set it to 15 GB for starters, and it was enough.

sudo mount -o remount,size=15G,noatime /run/user/1000

4. Adjust swap space

4.1 Temporarily

Used this Askubuntu answer in the following way:

Check current swap:

$ free -th
              total        used        free      shared  buff/cache   available
Mem:           7.7G        4.6G        253M        985M        2.8G        1.8G
Swap:          975M          0B        975M
Total:         8.6G        4.6G        1.2G

$ sudo swapon -s
Filename                                Type            Size    Used    Priority
/dev/dm-2                               partition       999420  3840    -2

Setting up the swap file, and turning it on:

$ sudo touch /temp_swap_15G.img
$ sudo fallocate -l 15G /temp_swap_15G.img
$ sudo mkswap /temp_swap_15G.img

# `-p` is the priority; the default is -2 and anything
# higher will be used first
$ sudo swapon -p 27 /temp_swap_15G.img

Checking the results:

$ sudo swapon -s
Filename                                Type            Size    Used    Priority
/dev/dm-2                               partition       999420  4352    -2
/temp_swap_15G.img                      file            15728636        0       27

4.2 Permanently

StackExchange answers (snapshots available on archive.org):

This is a troubleshooting one for the LVM one: