[Solved] Running into issues enforcing root disk size on containers

1

I’m just getting started with Inucs and have set up a single server that I’m managing with terraform.

The incus server runs an ext4 filesystem; nothing fancy for this lab test. Eventually I’d probably run it on zfs.

I’m trying to enforce a 16GB root filesystem limit on a simple 22.04 container that I created in terraform:

resource "incus_instance" "testinstance" {
(...)
  device {
    type = "disk"
    name = "root"
    properties = {
      pool = incus_storage_pool.default.name,
      path = "/",
      size = "16GiB",
    }
  }

This creates an incus instance on my server with the following:

$ incus config show testinstance

config:
  image.architecture: amd64
  image.description: Ubuntu jammy amd64 (20240215_07:42)
  image.os: Ubuntu
  image.release: jammy
  image.serial: "20240215_07:42"
  image.type: squashfs
  image.variant: cloud
(...)
  devices:
    eth0:
      nictype: bridged
      parent: bridge0
      type: nic
    root:
      path: /
      pool: default
      size: 16GiB
      type: disk

Here’s the default storage pool; very simple (on an ext4 filesystem that’s about 500GB):

$ incus storage show default

config:
  source: /var/lib/incus/storage-pools/default
description: ""
name: default
driver: dir
used_by:
- /1.0/instances/testinstance
status: Created
locations:
- none

Here’s the problem I’m running into

If I run df -h on the created testinstance container, I see the full filesystem size of the parent incus server:

$ incus exec testinstance -- df -h

Filesystem      Size  Used Avail Use% Mounted on
/dev/sda1       495G  8.5G  487G   2% /
(...)

And I’m able to write test files from within the container that exceed 16GB and end up taking up the entire parent server’s disk.

Is there something I missed with setting the 16GB limit?

2

Hmm, maybe this actually is an issue with ext4. I just tried making another storage pool that’s backed by btrfs and, although I still see the parent incus server’s free disk space, I do see quotas being enforced:

root@testinstance:~# fallocate -l 100G test
fallocate: fallocate failed: Disk quota exceeded
root@testinstance:~# ls
test
root@testinstance:~# df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/sdb        512G  749M  510G   1% /

This is on incus 0.5.1 running on a dev build of 24.04 with kernel 6.6.0-14-generic

3

Yeah, I think I’ll call this one solved for now unless there’s a known workaround or there’s something that I did wrong above.

I also came across User quota support on ext4/lvm root volume which looks related.

I’ve wanted to move ahead with zfs and/or btrfs anyways