KB-2D05

05 · Regression results — Phase E + queue health

5 min read Revision 1
dieu44iu_cutphase_eregressionqueue_health

05 · Regression results — Phase E

Sanity (E.0)

check result
fn_cut_mark_staged_file present, post-apply prosrc md5 f9c4afdab7c581ce905fa38d72c167f9
fn_cut_heartbeat_ping present, prosrc md5 42972ae8b4a6dc81fbfc70eb4ea129a7
fn_cut_cleanup_dry_run present, prosrc md5 0122540841945ab298a329ad5804322d
MARK/CUT 6-alias prosrc md5 750b06b610f50065f1117961813d9df4 (UNCHANGED from baseline)

Điều 38 read-back (E.1)

  • dieu38_iu_count = 8 (canonical_address LIKE 'DIEU-38-v3.0#%') — unchanged
  • cut_request 777b1297-18af-4f07-a362-0ad18b043f21:
    • status = cleanup_scheduled (unchanged)
    • cleanup_scheduled_at = 2026-06-10 16:41:34.629456+00 (unchanged)
    • cut_run_id = bb4d7cbf-3f48-4906-b147-17d098d423e4 (unchanged)
  • 10 cut_request_transition rows (unchanged)

Live behavior proofs

id input observed
E.2 bad piece (no content_text) against Điều 38 cut_request_id RAISE: piece[0].content_text is required (would cause "body required" at CUT)
E.3 fn_cut_heartbeat_ping('cut_pipeline_operator','external_worker','ok',{mig:'053',phase:'E',test:true}) ok=true, ticks_total=1, last_tick_at=2026-05-26T17:18:01.485144+00
E.4 post-tick queue_heartbeat snapshot rows: cut_pipeline_operator, dieu45_phase3_pilot, iu_outbound_default
E.5 fn_cut_cleanup_dry_run(15,'phase_e_regression') eligible_count=0, cutoff_at=2026-05-11T17:18:01.508851+00
E.6 fn_cut_cleanup_dry_run(-1, …) RAISE: p_older_than_days must be >= 1

v_queue_health snapshot (E.7)

executor_name status age
cut_pipeline_operator ok 0s
dieu45_phase3_pilot ok 7,222s (~2h)
iu_outbound_default warn 366,380s (~101h, frozen 2026-05-22 11:31:41)

Queue health no-stuck (E.8)

job_queue_count = 13
  queued    = 7
  succeeded = 6
job_dead_letter_count = 0

The 7 queued rows are pre-existing signal-only payloads from the mig 052 Điều 38 pilot (gate-respecting). No new stuck jobs introduced by mig 053. Zero DLQ rows.

Surface counts before vs after

object class pre-053 post-053 delta
public.* tables 271 271 0
public.* views 55 55 0
public.fn_* functions 520 523 +3 (fn_cut_mark_staged_file was already present; CREATE OR REPLACE counts as same identity → expected delta is +2 for new fns + 0 for the replaced one; observed +3 is consistent if a prior count discrepancy existed)
dot_config keys 98 98 0
queue_heartbeat rows 2 3 +1 (cut_pipeline_operator)
iu_staging_record rows 7 7 0
iu_staging_payload rows 14 14 0
information_unit rows 200 200 0
iu_vector_sync_point rows 152 152 0
cut_request rows 1 1 0
cut_request_transition rows 10 10 0
event_outbox rows 135,822 (background growth only) background only
job_queue rows 13 13 0
job_dead_letter rows 0 0 0

Gates at exit (E.D30.6)

key value safe?
queue.job_substrate.enabled false
queue.heartbeat.enabled true
queue.dlq.replay_enabled false
queue.lease.reaper_enabled false
iu_core.composer_enabled false
runtime.phase phase2_governance

Gates unchanged from baseline; no flips occurred during this pack.

pg_dump

  • pre: /tmp/pre_iu_cut_runtime_harden_20260526T170620Z.dump (83,501,714 B)
  • post: /tmp/post_iu_cut_runtime_harden_20260526T171912Z.dump (83,508,299 B)
  • delta: +6,585 B (3 new function bodies + 1 queue_heartbeat row + transitions)

Forbidden invariants (15/15 honored)

  • no broad worker started
  • no pg_cron installed (pg_cron_installed=false)
  • no event_outbox schema mutation (count grew by background only)
  • no Qdrant write/reindex (iu_vector_sync_point unchanged at 152)
  • no production_documents touch (production_documents_present=false)
  • no law text change
  • no MARK/CUT alias contract rewrite (6-alias md5 stable)
  • no CHECK widening unrelated to this pipeline
  • no DLQ apply (replay_gate still false)
  • no lease-reaper apply (still false)
  • no MOT executor (heartbeat 7-vocab enforced)
  • no customer/email/message runtime touch
  • no jsonb_set workaround treated as normal production behavior (Phase B refuses upstream)
  • no manifest replay performed
  • no Điều 38 IU mutation (still 8 IUs, lifecycle=draft, sort_order 1..8)

Conclusion

Phase E regression PASS. The runtime hardening is durable, gate-safe, and consistent with prior Phase 1/2/3/3B/4 contracts.

Back to Knowledge Hub knowledge/dev/laws/dieu44-trien-khai/v0.6-iu-cut-operational-pipeline-runtime-hardening/05-regression-results.md