Closed rvagg closed 2 years ago
Base: 14.04% // Head: 13.98% // Decreases project coverage by -0.06%
:warning:
Coverage data is based on head (
a038aaf
) compared to base (6a312a1
). Patch coverage: 0.00% of modified lines in pull request are covered.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
I'll wait for @hannahhoward for this one because I don't know how to deploy this to the bedrock kubernetes - I'd like to know though.
We're seeing a lot of entries in the event recorder DB for successful queries that don't go to retrievals (for any SP), and at the same time for some SPs we're seeing excessive numbers of queries. I believe this is simply because we're already at max concurrent retrievals for that SP and we're performing queries but not getting to do a retrieval for them because we only check concurrent count once we get to retrieval phase.
So this change will short-circuit an SP's inclusion in the candidate list if we're already at max concurrent retrievals for that SP. It's not always going to be the case that if we're at max-concurrent that we won't be able to retrieve, there's a small window where a retrieval could finish before we finish query phase; but it's more likely that we'll just end up bothering SPs with excessive pointless queries that don't result in a retrieval.