QA » openQA Project » openQA Tests » qe-yam

See https://jira.suse.com/browse/SLE-17081

https://trello.com/c/Ow3Vlrmx/1734-2-second-part-of-sle-17081-unification-of-yast2-bootloader-udev-code-try-again

https://trello.com/c/L8Xiojzb/2215-3-first-part-of-sle-17081-enforce-uuid-for-new-swap-partitions

Part for swap partitions was tested by riafarov for other feature. fstab entries have to be still validated.

As always, proposal for automation should be files as a ticket for qe-yast project.

Content of trello cards:

We are using [SLE-17081](https://jira.suse.com/browse/SLE-17081) as an opportunity to bring this card back to $CURRENT (we wanted to do it for SLE-15-SP3 anyways).

This card was originally created as a follow-up of https://trello.com/c/SFYLkAIX/, that was a follow-up/consequence of https://trello.com/c/TyCrUgrP/

yast-bootloader contains logic for finding the preferred (most stable) device name to reference a device. See https://github.com/yast/yast-bootloader/blob/master/src/lib/bootloader/udev_mapping.rb#L47 and implementation at https://github.com/yast/yast-bootloader/blob/master/src/lib/bootloader/udev_mapping.rb#L110

It makes sense for yast-bootloader to just rely on yast-storage-ng instead of implementing the logic on its own. Of course, the yast-storage-ng API may need some refinements to make it possible.

Tip: `MountByType.best_for` contains the related logic in yast-storage-ng

See also https://bugzilla.suse.com/show_bug.cgi?id=1169874.

## Testing opportunity

Let's use this as an opportunity to manually test the FULL installation process to check how this works together with:

- https://trello.com/c/L8Xiojzb/
- https://github.com/yast/yast-yast2/pull/1126 - [bsc#1169874](https://bugzilla.suse.com/show_bug.cgi?id=1169874)
- https://github.com/openSUSE/libstorage-ng/pull/790 - [bsc#1180560](https://bugzilla.suse.com/show_bug.cgi?id=1180560)


## Review

- PR for storage-ng (includes a lot of info): https://github.com/yast/yast-storage-ng/pull/1196
- PR for bootloader: https://github.com/yast/yast-bootloader/pull/628
- Logic to determine the better available name moved to `BlkDevice#preferred_name` and `BlkFilesystem#preferred_name`.
- Tested manually (with a dud including both PRs and the most recent libstorage-ng, and yast2.rpm)
  - Full installation of Tumbleweed with role XFCE and default settings (verified everything works and it includes the `resume=/dev/disk/by-uuid/xxyyzz` parameter)
  - Full installation of Tumbleweed with role XFCE changing default mount_by to path (verified everything works and it includes the `resume=/dev/disk/by-path/aa:bb-cc` parameter)

See [SLE-17081](https://jira.suse.com/browse/SLE-17081), which is a follow-up of [bug#1177926](https://bugzilla.suse.com/show_bug.cgi?id=1177926), which is basically the same as [bsc#1169874](https://bugzilla.suse.com/show_bug.cgi?id=1169874).

The problem is about using "resume=/dev/disk/by-id/$whatever" in the bootloader line. The bootloader proposal indeed tries to use UUID in most situations. See https://github.com/yast/yast-bootloader/blob/master/src/lib/bootloader/udev_mapping.rb#L47

But the problem is that, in most cases, the swap device is still not created in the moment the bootloader proposal is calculated. Thus the UUID is not known and the bootloader proposal falls back to another identification method.

This card is about imply enforcing an UUID for all new swap devices, instead of delegating the generation of the UUID to mkfs.

### Review

- https://github.com/yast/yast-storage-ng/pull/1192 (includes description and screenshot)
- Other problems found during manual testing:
  - [bug#1180218](https://bugzilla.suse.com/show_bug.cgi?id=1180218): bootloader proposal cannot handle changes in the swap device
  - [bug#1180222](https://bugzilla.suse.com/show_bug.cgi?id=1180222): reusing UUIDs of deleted swaps does not work in all cases