For those of you running into the issue, can you include your Sidekiq logs (/var/log/gitlab/sidekiq/current) and see if there are any errors with RebaseWorker?
@jamedjo As for https://gitlab.com/gitlab-com/runbooks/merge_requests/286, I see that branch is behind by a few commits. The current logic checks that that the merge request has a direct ancestor to the target branch. Does clicking on the Rebase button fix the problem?
@markglenfletcher I don't think it actually is a duplicate - the problem there is that when there is a 'legitimate' error, nothing is shown; and the problem here is that PostReceive wasn't working (which fixing #1774 (closed) wouldn't necessarily help with). However, it seems to be solved, so I will close it.
@felix.meyner This looks like a different problem. It says "the rebase button is always there even if the branch is up to date against master. The merge button is invisible." Although the animation on the report does look similar to #2979 (closed), however, it seems the problem in this issue is that the rebase button is presented even though no rebasing should be required.
@david.grant Maybe the rootcause is the same? In your ticket the you cannot fast forward merge.
Here we could because we have a clean history
But in both cases the result of the rebase (an error or the positive response) is not processed by the frontend in both tickets.
If you close the MR and reopen it, it should reset the HEAD SHA we store for the MR, which will also make it realise the history is linear. I'm sorry that we broke this again - let's keep tracking the root cause in https://gitlab.com/gitlab-org/gitlab-ee/issues/2979