tests: create partition scheme test

[lukewarmtemp]

Create a test to verify partition numbers and sizes. This is to ensure that the partition scheme is valid during the transition to using OSBuild in the FCOS pipeline.

Jira Issue: https://issues.redhat.com/browse/COS-2527

[dustymabe]

we might need to add some special consideration here for 4k disks, though hmm, do these tests even run for 4k disks or is that just testiso tests?

[lukewarmtemp]

we might need to add some special consideration here for 4k disks, though hmm, do these tests even run for 4k disks or is that just testiso tests?

Oh it didn't even cross my mind to test the different platforms. I just ran it on the qcow2 built by OSBuild and COSA, but I can try running it for different platforms and see what the outputs are.

[dustymabe]

we might need to add some special consideration here for 4k disks, though hmm, do these tests even run for 4k disks or is that just testiso tests?

Oh it didn't even cross my mind to test the different platforms. I just ran it on the qcow2 built by OSBuild and COSA, but I can try running it for different platforms and see what the outputs are.

Like I said 4k disks probably won't be able to be tested with this because I think the only place we can use those is for bare metal and we only test the bare metal images using our testiso stuff.

But... having this test run everywhere (like AWS, GCP, etc) is useful, so making it a test that isn't limited to qemu will have value I think.

[lukewarmtemp]

But... having this test run everywhere (like AWS, GCP, etc) is useful, so making it a test that isn't limited to qemu will have value I think.

Test runs on AWS successfully.

[lukewarmtemp]

A point of investigation: the qcow2 created by OSBuild does not have the partlabel field

lsblk --json --bytes -o NAME,FSTYPE,START,SIZE,MOUNTPOINT,PARTLABEL,PARTTYPENAME,LOG-SEC "$disk_name" | jq '.blockdevices[]'
{
  "name": "vda",
  "fstype": null,
  "start": null,
  "size": 10737418240,
  "mountpoint": null,
  "partlabel": null,
  "parttypename": null,
  "log-sec": 512,
  "children": [
    {
      "name": "vda1",
      "fstype": null,
      "start": 2048,
      "size": 1048576,
      "mountpoint": null,
      "partlabel": null,
      "parttypename": "BIOS boot",
      "log-sec": 512
    },
    {
      "name": "vda2",
      "fstype": "vfat",
      "start": 4096,
      "size": 133169152,
      "mountpoint": null,
      "partlabel": null,
      "parttypename": "EFI System",
      "log-sec": 512
    },
    {
      "name": "vda3",
      "fstype": "ext4",
      "start": 264192,
      "size": 402653184,
      "mountpoint": "/boot",
      "partlabel": null,
      "parttypename": "Linux filesystem",
      "log-sec": 512
    },
    {
      "name": "vda4",
      "fstype": "xfs",
      "start": 1050624,
      "size": 10199481856,
      "mountpoint": "/sysroot",
      "partlabel": null,
      "parttypename": "Linux filesystem",
      "log-sec": 512
    }
  ]
}

[dustymabe]

A point of investigation: the qcow2 created by OSBuild does not have the partlabel field

yep. definitely something that may be causing us problems. We'll need to remedy that I think.

[dustymabe]

I think this LGTM - can we fix the CI failures? Once we do that I'll give it a spin on a 4k disk just to verify it works (I know it won't ever get run against a 4k disk, but in theory it should just work).

[lukewarmtemp]

can we fix the CI failures

Not really sure why it's failing... I used truncate -s -1 to get rid of the whitespace at the end of the file.

[dustymabe]

let's squash this into one commit?

[dustymabe]

LGTM