Add secondary search queue #6010
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Currently, the search permit provider handles only one query at a time. This is good for the overall query latency when queries are relatively short. It is catastrophic in any concurrent setup where some queries are slow, because all the other queries will be starved from resources until the large query completes.
This PR creates a secondary search permit provider that is independently configured. "Fast" leaf queries are routed to one provider and "slow" ones to the other. Query duration is naively inferred from the number of targetted splits, and the threshold for being routed to one provider or the other can be configured using
SearcherConfig.secondary_targeted_split_count_threshold.One limitation of the current implementation of this PR is that for a given root search, various leaf searches might fall on different sides of the threshold. We could fix this by applying the threashold on the root and add the targeted provider in the leaf request.
How was this PR tested?
We ran this setup on our highly concurrent production environment. It helped stabilize query performance drastically.