TEST ONLY: SymDB extractor without throttle (regression detection check for #5776)#5784
TEST ONLY: SymDB extractor without throttle (regression detection check for #5776)#5784p-datadog wants to merge 1 commit into
Conversation
|
👋 Hey @DataDog/ruby-guild, please fill "Change log entry" section in the pull request description. If changes need to be present in CHANGELOG.md you can state it this way **Change log entry**
Yes. A brief summary to be placed into the CHANGELOG.md(possible answers Yes/Yep/Yeah) Or you can opt out like that **Change log entry**
None.(possible answers No/Nope/None) Visited at: 2026-05-19 22:18:21 UTC |
|
DO NOT MERGE. Sibling-branch test for PR #5776: removes the SLEEP_EVERY_N_MODULES / SLEEP_SECONDS throttle from collect_extractable_modules to verify that the benchmarks added in #5776 detect the regression bp-runner is supposed to gate. Expected signals on CI: - symbol_database_background_impact: p99 ratio treatment_over_baseline jumps from ~1.05 to >1.5 - symbol_database_baseline_matrix: 100%-arm rps_drop jumps from ~8% to ~25%+; lower-baseline arms may also shift - bp-runner: significant ops/sec drop on workload arms of both benchmarks vs master (master here = PR #5776 tip with throttle intact) If the throttle is later restored to master, this branch can be deleted.
3dd7ca2 to
c70a393
Compare
Typing analysisNote: Ignored files are excluded from the next sections.
|
JIRA: DEBUG-5668
Summary
Do not merge. This is a sibling-branch test for #5776.
Branches from #5776's tip and reverts only the throttle implementation in
lib/datadog/symbol_database/extractor.rb+sig/datadog/symbol_database/extractor.rbs. The benchmarks and helper additions from #5776 are kept.The point is to verify that the benchmarks added in #5776 actually catch a throttle regression, by running CI against a branch where the throttle has been removed.
What's removed vs #5776's tip
SLEEP_EVERY_N_MODULES = 100constantSLEEP_SECONDS = 0.001constantprivate_constant :SLEEP_EVERY_N_MODULES, :SLEEP_SECONDSsleep SLEEP_SECONDS if (seen % SLEEP_EVERY_N_MODULES).zero?call (and itsseencounter) incollect_extractable_modulesWhat to check in CI
Compare CI output here vs #5776:
symbol_database_background_impact—p99_ratio_treatment_over_baselineshould jump from ~1.05 (PR 5776 with throttle) to well above 1.5 (this PR). The in-scriptenforce_requirementgate at threshold 1.50 should fireexit 1.symbol_database_baseline_matrix— the 100%-armrps_dropshould jump from ~8% to materially higher. Lower-baseline arms are noisier in microbenchmark form (documented limitation), but should also shift.symbol_database_extraction—wall_time_secondsshould drop ~2x (no per-chunk sleep) andcpu_percentshould approach 100%. Memory should be unchanged.Cleanup
Delete this branch once the comparison has been recorded.
Change log entry
None.