test(storage#812): regression cover for 2.1.24 invocation-bookend gap (#813 follow-up)#814
Merged
Merged
Conversation
#813 switched 2.1.24 checkpointingTimestampGap from the DLIO summary start/end to the invocation bookends (metadata.invocation_start_time / invocation_end_time), but shipped without a test — the existing 2.1.24 cases all pass metadata={} and exercise only the summary fallback, leaving the new bookend behavior uncovered. Add three cases to the 2.1.24 sentinel file: - gap_uses_invocation_bookends_not_summary: the #812 scenario — a valid pair whose true quiet window is 5 s but whose DLIO summary window is inflated to a 65 s apparent gap by read startup + write cluster collection. Must pass. (Verified RED against the pre-#813 summary-based code, GREEN with #813.) - bookend_gap_still_flags_a_genuine_long_pause: a 5-minute bookend gap against 10 s invocations must still fail — the source-swap must not neuter the check. - falls_back_to_summary_when_bookends_absent: pre-bookend results dirs (metadata without the keys) still validate via the summary fallback. Test-only; the fix landed in #813.
|
MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅ |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Regression coverage for the 2.1.24
checkpointingTimestampGapfix that #813 (fixing #812) shipped without a test.Why
#813 correctly switched 2.1.24 from the DLIO summary
start/endto the invocation bookends (metadata.invocation_start_time/invocation_end_time), mirroring 4.7.1cache_flush_validation. But it added no test, and the existing 2.1.24 sentinel cases all passmetadata={}— so they exercise only the summary fallback path, leaving the new bookend behavior (the entire point of the fix) uncovered. A future refactor could silently regress it.Tests added (test-only — no source change; the fix is in #813)
Three cases in the existing 2.1.24 sentinel file:
gap_uses_invocation_bookends_not_summary— the False2.1.24 checkpointingTimestampGapviolation (same issue of flush timing calcuation) #812 scenario: a valid write+read pair whose true quiet window is 5 s, but whose DLIO summary window is inflated to a 65 s apparent gap by read-side framework startup + write-side cluster collection. Must pass. Verified RED against the pre-Implement correct calculation for checkpointingTimestampGap #813 summary-based code and GREEN with Implement correct calculation for checkpointingTimestampGap #813, so it genuinely guards the fix.bookend_gap_still_flags_a_genuine_long_pause— a 5-minute bookend gap against 10 s invocations must still fail 2.1.24; proves the source-swap did not neuter the check.falls_back_to_summary_when_bookends_absent— results dirs predating the bookends (metadata without the keys) still validate via the summary fallback.Testing
Full CI-equivalent sweep, all green:
mlpstorage_py/tests885 passed (882 + 3 new),tests2930 passed (1 skipped),vdb_benchmark/tests228 passed,kv_cache_benchmark/tests238 passed. Based onmainafter #813 merged, so CI is green.