Pinging @elastic/es-core-features

@DaveCTurner I agree that would be better, however it looks like a fair bit of plumbing work to get that information through the NodeInfo API (or modify the deprecation check framework to support getting information another way). Am I wrong there?

This is an interesting one, as there doesn't look like a clean way to expose this easily.

Pinging @elastic/es-distributed

I see. Yes, you're right, apart from logging a deprecation warning we don't record that the discovery file isn't in the new place.

I think if the file is not in the right place on 7.0+ nodes during an upgrade then the upgraded nodes will struggle to join the cluster, which seems ok. It depends a bit on whether they're using file-based discovery exclusively or some combination of other discovery mechanisms too, but if it's exclusively file-based discovery then no file means no discovery. Am I right that the migration assistant is intended to be a best-effort thing? If so then this is going to be ok.

@DaveCTurner I believe that yes, the migration assistant/deprecation info API is intended to be best-effort. Users will still need to check the deprecation logs for any request/response format changes they'll need to make anyway.

I've opened #36343 to track the larger issue, but it's unlikely to be addressed before the release of 7.0, so I think it might be best to move forward with this as-is.