Currently in chromite:

use_lkgm is unused (we appear to have eliminated its use here: https://chromium-review.googlesource.com/#/c/307186/)

use_chrome_lkgm is only used for chrome_perf which is not on any waterfall.

The current failure is just a misconfiguration on the bot. It's using chrome-bot@chromium.org but it should be using chromeos-commit-bot@chromium.org.

remote: ERROR:  In commit fc4fbedda3423ed1c9e2fa9bd79721dffbe6c5f8        
remote: ERROR:  committer email address chrome-bot@chromium.org        
remote: ERROR:  does not match your user account.        
remote: ERROR:        
remote: ERROR:  The following addresses are currently registered:        
remote: ERROR:    chromeos-commit-bot@chromium.org        
remote: ERROR:        
remote: ERROR:  To register an email address, please visit:        
remote: ERROR:  https://chromium-review.googlesource.com/#/settings/contact        

In general (copy-pasted from previous emails on the topic):

The Chrome LKGM is used for ensuring that the Chrome OS version approximately matches the Chrome version in Simple Chrome. While it's OK to have a little version skew, if your checkout has an old version of Chrome you don't want Chrome OS to be constantly updating unless you similarly are syncing your Chrome checkout. Too much version skew can definitely break things. The Chrome LKGM accomplishes its purpose, but it falls behind often because the canaries fail so often.

We do have the concept of a Chrome OS per-board LKGM, but it doesn't map to Chrome versions so it isn't usable for this purpose.

Given the purpose of the Chrome LKGM (I.e. just to map between Chrome versions and Chrome OS versions), it would actually be reasonable to just update it even if the canaries are failing, because we already have logic to look at the per-board LKGM in this case. That would fix the issue of the Chrome LKGM failing and would get people using newer builds (but only if the board they wanted the latest version of was passing). In this case, we'll likely need to rename the file from LKGM to something else.

If you wanted to design a fresh approach, you'd need to design something else that maps between checkouts of Chrome and the binary distributions of Chrome OS: For a given version of Chrome, what Chrome OS version should you use?

reassigning to  stevenjb@ for input based on david's comment.

This builder is still a pita to maintain, but I don't think we have the cycles to replace it any time soon. At least it is better documented in YAQS now:

https://yaqs.googleplex.com/eng/q/5254238507106304

Resolving WontFix.