Test raw disk deploy fails

[mudler]

Seems the test raw disk deploy started to fail recently:

• Failure [165.664 seconds]
cOS Recovery deploy tests
/Users/runner/work/cOS-toolkit/cOS-toolkit/tests/recovery-raw-disk/recovery_raw_disk_test.go:9
  after running recovery from the raw_disk image
  /Users/runner/work/cOS-toolkit/cOS-toolkit/tests/recovery-raw-disk/recovery_raw_disk_test.go:23
    uses cos-deploy to install [It]
    /Users/runner/work/cOS-toolkit/cOS-toolkit/tests/recovery-raw-disk/recovery_raw_disk_test.go:24

    Unexpected error:
        <*errors.withStack | 0xc0000b0d80>: {
            error: <*errors.withMessage | 0xc0000d92c0>{
                cause: <*ssh.ExitError | 0xc0000f1a80>{
                    Waitmsg: {status: 1, signal: "", msg: "", lang: ""},
                },
                msg: "\x1b[36mINFO\x1b[0m[2022-03-24T01:05:53Z] Starting elemental version v0.0.13           \nError: no device found\n\x1b[31mERRO\x1b[0m[2022-03-24T01:05:54Z] State partition 'COS_STATE' not found        \n",
            },
            stack: [0x12cefdd, 0x12d129a, 0x12d1280, 0x126d27a, 0x126cc45, 0x126c33b, 0x126f4a9, 0x126ee87, 0x1284e85, 0x1284ba5, 0x12843e5, 0x1286692, 0x1291ca5, 0x129196b, 0x12d0f65, 0x10dc202, 0x1069ca1],
        }
        INFO[2022-03-24T01:05:53Z] Starting elemental version v0.0.13           
        Error: no device found
        ERRO[2022-03-24T01:05:54Z] State partition 'COS_STATE' not found        
        : Process exited with status 1
    occurred

    /Users/runner/work/cOS-toolkit/cOS-toolkit/tests/recovery-raw-disk/recovery_raw_disk_test.go:28
------------------------------

Additional context

• https://github.com/rancher-sandbox/cOS-toolkit/runs/5668948084?check_suite_focus=true
• https://github.com/rancher-sandbox/cOS-toolkit/runs/5668591417?check_suite_focus=true
• https://github.com/rancher-sandbox/cOS-toolkit/runs/5668740862?check_suite_focus=true
• https://github.com/rancher-sandbox/cOS-toolkit/runs/5684036928?check_suite_focus=true

[davidcassany]

Found the bug. Even I can't explain it did not pop up until now. The issue is related to the partitioner or, more specific, to the yip plugin that makes use of the partitioner. I am about to fix it, the is issue is that parted does not extend the partition table if it does not match with the current device (some image got dd'ed to a bigger disk), in such case it throws an error when creating a new partition unless it is fixed before hand. This is something I missed to port on the elemental reimplementation of the plugin.

[davidcassany]

This is uglier than what I initially thougth, it strikes back to the sgdisk vs parted. parted can't fix the GPT header to match disk size in non-interactive mode (yes, unbelievable I found references to this up to 2015 and it is still an issue). So the solution is using sgdisk only for this specific use case or alternatively assume some interaction on parted side and pipe expected input. 😞

[davidcassany]

Requires the elemental bump still