mariadb: log primary reconcile drift predicates

[weicao]

Problem

Fresh focused r11 on addon combined 73d05dcfd2a907b7bbcb292d2f0d8d3de7d7abf1 and syncer 7502b58f45efd4fe8f2b06e5313230d9aec69510 proved an addon-side marker oscillation. After an authoritative primary write commit succeeds, the runtime primary listener reconciler enters primary-role-state-drift and starts repair before the next HA promotion. Repair retracts the authority markers, but the current aggregate log does not identify which of its four predicates first failed.

Evidence:

• r11 first-red outer SHA256: 5ca8e2cc2a71c74d177ef3b8c4e0dd2633bab90e75a3320964494130ce6f579a
• r11 first-red manifest SHA256: cf3d4b219a8ad4416b9133b3a615f787a86d30389677bf88fa68c12bd83081fc (14/14)
• supplemental addon-log outer SHA256: 83aabd76e9612f9291bdefae358f75e2bc4622a5b35955f69ed00299eee69096
• supplemental manifest SHA256: 234e06383a2362f02bbd2a5fff82cab5a0c27177ec82dd10d7d5aae345a5ee60 (3/3)
• observed sequence: authority commit/reconcile complete at 03:22:20Z; addon runtime-primary-listener-reconcile-repair-begin at 03:22:24Z; only afterwards does the repair fail closed.

Change

Capture and log the four primary-publish predicates before repair can mutate state:

• primary_ready
• remote_root_fence
• master_info
• listener_wildcard

The fast-path decision uses that same captured snapshot. This is diagnostic observability only; it does not change the repair policy or claim to fix the oscillation.

TDD and source gates

• RED: focused ShellSpec 30 examples, 1 failure before production instrumentation
• GREEN: focused ShellSpec 30 examples, 0 failures
• full MariaDB ShellSpec: 768 examples, 0 failures, 9 historical pendings
• bash -n addons/mariadb/scripts/replication-entrypoint.sh
• shellcheck -S error addons/mariadb/scripts/replication-entrypoint.sh
• helm lint addons/mariadb
• default helm template plus rendered predicate-log verification
• git diff --check

Runtime boundary

The next focused round is diagnostic N=1 only. Its acceptance target is an unambiguous pre-mutation value for all four predicates across the commit-to-repair window. It is not release evidence, and the full suite remains NO-LAUNCH.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 0% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 0.00%. Comparing base (73d05dc) to head (3e51be9).

Files with missing lines	Patch %	Lines
...plication_merged_semisync_startup_recovery_spec.sh	0.00%	21 Missing ⚠️

Additional details and impacted files

@@                                 Coverage Diff                                  @@
##           helen/mariadb-open-prs-combined-20260715t030336z-r16   #3214   +/-   ##
====================================================================================
  Coverage                                                  0.00%   0.00%           
====================================================================================
  Files                                                       154     154           
  Lines                                                     24252   24270   +18     
====================================================================================
- Misses                                                    24252   24270   +18     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.