Conversation
…L-string sniff derive_agg_role() used to guess a workload entry's AggRole (Quantile/ Sum/Count/Topk/Other) by sniffing the leading token of its PromQL query_string -- the same duplicate-classifier smell already retired from the live serving path (the old two-analyzer comparison in analyzer-parity-matrix.md, #422). Now parses the query through the same canonical pipeline capability_for()/serving uses (query_parser::parse_query_expr_canonical), runs it through the L3 rule-based optimizer (QueryOptimizer::new(0.0).optimize -- TopKFusion is a pure structural rewrite, ignores the cost model, so the placeholder throughput doesn't affect the outcome) so topk(k, m) actually reaches an Aggregate{TopK} node instead of staying Sort+Limit, then classifies by the real outer AggIntent (collect_agg_intents, now pub(crate) for this reuse). Two things this surfaced: - `(quantile_over_time(0.9, m[5m]))` (redundant wrapping parens) -- the old leading-token sniff finds an empty token at a `(` and mis-defaults to Sum; real classification is unaffected by surface punctuation. New test pins this. - `topk_over_time(...)` was in the old test's query list but isn't a real function this parser (or vanilla PromQL) recognizes at all -- confirmed via grep, nowhere in query_parser/promql.rs or intent_algebra/lower.rs. A workload entry with that query_string would fail to parse anywhere else in the real pipeline too, so the old heuristic classifying it as Topk was itself the bug. Test updated to drop it, with an explanatory comment. cargo build --workspace clean; control_plane 727/727 (one pre-existing, unrelated skip as before), workload:: 33/33 (32 + 1 new test). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Merged
3 tasks
Contributor
Author
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.
Summary
derive_agg_role()used to guess a workload entry'sAggRole(Quantile/Sum/Count/Topk/Other) by sniffing the leading token of its PromQLquery_string— the same duplicate-classifier smell already retired from the live serving path (the old two-analyzer comparison inanalyzer-parity-matrix.md, #422). Now parses the query through the same canonical pipelinecapability_for()/serving uses (query_parser::parse_query_expr_canonical), runs it through the L3 rule-based optimizer (QueryOptimizer::new(0.0).optimize—TopKFusionis a pure structural rewrite, ignores the cost model, so the placeholder throughput doesn't affect the outcome) sotopk(k, m)actually reaches anAggregate{TopK}node instead of stayingSort+Limit, then classifies by the real outerAggIntent(collect_agg_intents, nowpub(crate)for this reuse).Two things this surfaced:
(quantile_over_time(0.9, m[5m]))(redundant wrapping parens) — the old leading-token sniff finds an empty token at a(and mis-defaults toSum; real classification is unaffected by surface punctuation. New test pins this.topk_over_time(...)was in the old test's query list but isn't a real function this parser (or vanilla PromQL) recognizes at all — confirmed via grep, nowhere inquery_parser/promql.rsorintent_algebra/lower.rs. A workload entry with thatquery_stringwould fail to parse anywhere else in the real pipeline too, so the old heuristic classifying it asTopkwas itself the bug. Test updated to drop it, with an explanatory comment.Test plan
cargo build --workspace— clean, no new warningscargo test --release --lib -p control_plane workload::— 33/33 (32 existing + 1 new)cargo test --release --lib -p control_plane— 727/727 (one pre-existing, unrelated skip as before)🤖 Generated with Claude Code