Comment 9 for bug 1847924

Robie, I've just updated the description with more information on the user impact (and why we should have this patch), and also the regression potential.

I wasn't completely right in my last comment, sorry - for RAID0/Linear arrays, the output of the array state in "mdadm --detail" may change after the patch _even for healthy arrays_. One example is the "readonly" state: we can write this state to the sysfs "array_state" file for a raid0/linear array, and currently the "mdadm --detail" will keep showing state "clean" (arguably an odd behavior, but it is how the state output works right now). With this patch, the state will show as "readonly".

I think we have two options here:

(a) We can backport the patch as is, hence risking some users running monitoring tools or any kind of scripts parsing "mdadm -detail" output breaking for raid0/linear arrays.

(b) We can introduce a conditional check for Bionic/Disco/Eoan in the "mdadm --detail" output to keep "clean" as the array state unless it's "active" or "broken" (the new state for failed arrays). Then, Focal will get a merge from a more recent version of mdadm with the original patch and diverge from this behavior, being aligned with upstream/other distros.

Basically, the discussion is to "when" introduce the user-visible change, if during the current releases or between releases (Focal being the first to change). There's an hypothetical 3rd alternative that I personally dislike and would avoid, that is introduce such conditional check for all Ubuntu releases from now on and diverge forever from upstream/other distros - I don't see a point in doing this.

Let me know your considerations Robie, and thanks for pointing out valid concerns!
Cheers,

Guilherme