Ubuntu-image command crashed

jamesh · March 29, 2021, 2:47am

I had a go at trying to implement the “cloud.conf in gadget snap” option, and am not having much luck. My model is set to grade: dangerous.

I can see that the cloud.conf from my gadget snap has been copied to /etc/cloud/cloud.cfg.d/80_device_gadget.cfg, but it isn’t obvious that any of the directives have been run (I tried using runcmd and write_files).

On the first boot (after the initial unpack/setup boot), the only reference to “cloud” I see in the journal is:

Mar 29 02:33:04 ubuntu snapd[1880]: devicemgr.go:891: System initialized, cloud-init reported to be in disabled state, disabled permanently

ogra · March 29, 2021, 9:36am

i cant find a referenc for this anymore (probably someone told it to me on IRC) but i happen to remember that the filename needs to be cloud.cfg in the gadget case and that you need a “nocloud” section in it (i could mis-remember or it might be outdated info tough) …

jamesh · March 29, 2021, 9:47am

I’d originally tried with cloud.cfg, which doesn’t work. When I switched to cloud.conf (as documented in @ijohnson’s post), I could see that the config file was being copied out of my gadget snap. And the snapd code is definitely looking for that file name:

github.com

snapcore/snapd/blob/4405f25a553d31876bfe5346948b9b8f0e2b67d4/sysconfig/cloudinit.go#L113-L121


	// next check if there is a gadget cloud.conf to install
	if HasGadgetCloudConf(opts.GadgetDir) {
		// then copy / install the gadget config and return without considering
		// CloudInitSrcDir
		// TODO:UC20: we may eventually want to consider both CloudInitSrcDir
		// and the gadget cloud.conf so returning here may be wrong
		gadgetCloudConf := filepath.Join(opts.GadgetDir, "cloud.conf")
		return installGadgetCloudInitCfg(gadgetCloudConf, WritableDefaultsDir(opts.TargetRootDir))
	}

I’ll have to look more into the nocloud thing. From a brief read of the docs, that seemed to be about feeding cloud-init data into the system from some external source. Here, the config has already been copied to /etc/cloud by snapd.

ijohnson · March 29, 2021, 11:40am

Can you share the cloud.conf you tried using? I think it’s a formatting issue in that it only works if you format the cloud.conf a certain way

jamesh · March 29, 2021, 1:42pm

This is the cloud.conf file I was using:

The write_files stanza is to a location that is made writable in the boot file system. The runcmd command references a script shipped inside one of the snaps required by the model definition.

ijohnson · March 29, 2021, 2:43pm

The cloud-init config that we have confirmed to work is formatted like this:

#cloud-config
datasource_list: [NoCloud]
users:
  - name: ubuntu
    sudo: "ALL=(ALL) NOPASSWD:ALL"
    lock_passwd: false
    ssh_authorized_keys:
      - "ssh-rsa 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 ian.johnson@canonical.com"

Perhaps you need to include datasource_list: [NoCloud] in your config? I don’t know how cloud-init internally decides if it’s used or not, but all snapd does to decide if cloud-init should be disabled or not is by querying if there is a datasource being used or not.

snowsky · March 29, 2021, 3:49pm

It works for me. A new user got created by my cloud.conf file.

jamesh · March 30, 2021, 3:47am

Yep. Adding datasource_list to my cloud.conf file seems to fix things. Both the write_files and runcmd directives seem to have been processed.

ijohnson · March 30, 2021, 2:40pm

Great thanks for confirming, I’ll make a note of this in my upcoming UC20 and cloud-init doc page

snowsky · July 14, 2021, 9:40pm

@ijohnson in the new version of UC, is there a plan to use the old way in uc18 to include the cloud.conf file when ubuntu-image command runs? Thanks.

ijohnson · July 14, 2021, 10:48pm

There is not currently a plan for that, no.

snowsky · July 15, 2021, 3:05pm

Thanks @ijohnson for the confirmation. When ubuntu-image runs, can we copy the file to ubuntu-seed directory, <ubuntu-seed>/data/etc/cloud/cloud.cfg.d/ with filenames ending with .cfg as you mentioned here, Ubuntu-image command crashed?

Cheers,

tobid · July 27, 2021, 12:43pm

Hi @ijohnson
I’m building an UC20 image at the moment and try to use a cloud-init config-file for some basic setup on a fresh install (adding some users, creating some files, etc.) as I did previously with a UC18 image.
For testing I basically went along this tutorial: https://ubuntu.com/core/docs/custom-images

Since I don’t want to create my own gadget.snap at the moment, I’m trying out your option 2, but it doesn’t seem to work for me. I build the UC20 image using ubuntu-image, then mount the ubuntu-seed partition of the finished image, create a directory <seed>/data/etc/cloud/cloud.cfg.d/ and copy a simple cloud.cfg (I tried cloud.conf as well) in there:

#cloud-config
datasource_list: [ NoCloud, None ]
users:
  - name: ubuntu
    homedir: /home/ubuntu
    sudo: ALL=(ALL) NOPASSWD:ALL
    groups: users, admin
    shell: /bin/bash

chpasswd:
  list: |
    ubuntu:ubuntu
  expire: False

My problem is, that this file doesn’t seem to be used when I later boot up the image (no user is created) and I’m not sure where to look for log-files in UC20.

Also the table you wrote into post Ubuntu-image command crashed confused me a bit. Is it possible the “Methodes” in this table are in the wrong order?

ijohnson · July 27, 2021, 9:30pm

@tobid can you show the model assertion for your image?

tobid · July 30, 2021, 7:28am

@ijohnson sure (just omitted my IDs):

{
  "type": "model",
  "series": "16",
  "base": "core20",
  "model": "my-pc",
  "architecture": "amd64",
  "authority-id": "...",
  "brand-id": "...",
  "timestamp": "2021-07-26T10:51:52+00:00",
  "grade": "signed",
  "snaps": [
    {
      "name": "pc",
      "type": "gadget",
      "default-channel": "20/stable",
      "id": "UqFziVZDHLSyO3TqSWgNBoAdHbLI4dAH"
    },
    {
      "name": "pc-kernel",
      "type": "kernel",
      "default-channel": "20/stable",
      "id": "pYVQrBcKmBa0mZ4CCN7ExT6jH8rY1hza"
    },
    {
      "name": "core20",
      "type": "base",
      "default-channel": "latest/stable",
      "id": "DLqre5XGLbDqg9jPtiAhRRjDuPVa5X1q"
    },
     {
      "name": "snapd",
      "type": "snapd",
      "default-channel": "latest/stable",
      "id": "PMrrV4ml8uWuEUDBT8dSGnKUYbevVhc4"
    },
    {
      "name": "htop",
      "type": "app",
      "default-channel": "latest/stable",
      "id": "hJmReLmgXSUj4SF7WhyTVRV6IzUa4QUZ"
    }
  ]
}

ijohnson · August 3, 2021, 2:17pm

In order to use cloud.conf on ubuntu-seed with UC20, you need to have grade: dangerous in your model assertion as indicated by the table above. The work for enabling this to work for grade: signed is not yet complete.

tobid · August 4, 2021, 8:38am

Thanks for the info, I’m going to try this.

I think I was a bit confused by that table. If I understand it right, I’m going for “Option 2” in your post which is described as

Put it in the ubuntu-seed partition at <ubuntu-seed>/data/etc/cloud/cloud.cfg.d/ with filenames ending with .cfg"

which says is allowed under these conditions:

Option 2 is allowed only with model assertion grade dangerous and signed (so this does not work for secured).

The table on the other hand has this option in the first line (named “Ubuntu-seed”, I guess), with different allowed grade. If this is a mistake in the post, maybe you could update it?

ijohnson · August 6, 2021, 12:38pm

sorry the option descriptions in that comment were out of order, I have fixed it now

jasonlo · April 21, 2022, 11:27am

I have recently hit some other issue when using ubuntu-image on 22.04 (jammy) lxd vm. The python3 version is 3.10.4

It is strange that it shows using …/214/usr/lib/python3.5/ Here is the error:

ERROR:ubuntu-image:COMMAND FAILED: sudo mount -oloop work/volumes/kria/part1.img /tmp/tmp7w2joncy/root-mount ERROR:ubuntu-image: ERROR:ubuntu-image:sudo: account validation failure, is your account locked? sudo: a password is required

ERROR:ubuntu-image:uncaught exception in state machine step: [11] populate_filesystems Traceback (most recent call last): File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/state.py”, line 82, in next step() File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/common_builder.py”, line 529, in populate_filesystems self._populate_one_volume(name, volume) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/common_builder.py”, line 477, in _populate_one_volume part.filesystem_label, preserve_ownership=True) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 256, in mkfs_ext4 with mount(img_file) as mountpoint: File “/snap/ubuntu-image/214/usr/lib/python3.5/contextlib.py”, line 59, in enter return next(self.gen) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 208, in mount run(‘sudo mount -oloop {} {}’.format(img, mountpoint)) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 123, in run proc.check_returncode() File “/snap/ubuntu-image/214/usr/lib/python3.5/subprocess.py”, line 659, in check_returncode self.stderr) subprocess.CalledProcessError: Command ‘[‘sudo’, ‘mount’, ‘-oloop’, ‘work/volumes/kria/part1.img’, ‘/tmp/tmp7w2joncy/root-mount’]’ returned non-zero exit status 1 ERROR:ubuntu-image:Crash in state machine Traceback (most recent call last): File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/main.py”, line 393, in main list(state_machine) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/state.py”, line 82, in next step() File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/common_builder.py”, line 529, in populate_filesystems self._populate_one_volume(name, volume) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/common_builder.py”, line 477, in _populate_one_volume part.filesystem_label, preserve_ownership=True) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 256, in mkfs_ext4 with mount(img_file) as mountpoint: File “/snap/ubuntu-image/214/usr/lib/python3.5/contextlib.py”, line 59, in enter return next(self.gen) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 208, in mount run(‘sudo mount -oloop {} {}’.format(img, mountpoint)) File “/snap/ubuntu-image/214/lib/python3/site-packages/ubuntu_image/helpers.py”, line 123, in run proc.check_returncode() File “/snap/ubuntu-image/214/usr/lib/python3.5/subprocess.py”, line 659, in check_returncode self.stderr) subprocess.CalledProcessError: Command ‘[‘sudo’, ‘mount’, ‘-oloop’, ‘work/volumes/kria/part1.img’, ‘/tmp/tmp7w2joncy/root-mount’]’ returned non-zero exit status 1

finish

local ret=1

‘[’ 1 -eq 0 ‘]’

echo ‘Build failed …’ Build failed …

Sorry if this is not relevant here, and if necessary I can create another issue. Thanks.

mborzecki · April 21, 2022, 11:33am

Looks like this is the reason.