"some time"? There is an ingest delay between the Butler emitting data and it getting loaded into LogDog. That delay is generally pretty short, but can be longer. If the stream doesn't exist in LogDog, there are two possibilities:
1) It will, but ingest delay. Maybe retry?
2) It will never exist, so retry is wasteful.
We could have Milo respond to 404s by retrying for some threshold (30s?) and then propagating the 404. Is there a better option here?
Comment 1 by estaab@chromium.org
, Feb 22 2017