From 8f860ab14d55d69987c3392d78d2fd4d8380b4fa Mon Sep 17 00:00:00 2001
From: admin <alec@bryanmail.net>
Date: Tue, 4 Aug 2026 20:04:50 -0700
Subject: [PATCH] postgresql-ha-final.yaml: add matching primary_slot_name
 (keep byte-identical with staging)

Mirrors the fix in postgresql-ha-staging.yaml: added
primary_slot_name: standby_leader_slot to both patroni-0 and patroni-1's
standby_cluster blocks so this file's specs stay byte-identical to
staging's, per this file's own "do not recreate the live promoted
primary/replica" design requirement. See postgresql-ha-staging.yaml
header and the ADR-0001 note for full incident detail.
---
 postgresql/cutover/postgresql-ha-final.yaml | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/postgresql/cutover/postgresql-ha-final.yaml b/postgresql/cutover/postgresql-ha-final.yaml
index b412dde..9744f51 100644
--- a/postgresql/cutover/postgresql-ha-final.yaml
+++ b/postgresql/cutover/postgresql-ha-final.yaml
@@ -1,4 +1,4 @@
-# ──────────────────────────────────────────────────────────────────────────
+# ─────────────────────────────────────────────────────────────────────────
 # postgresql/cutover/postgresql-ha-final.yaml — ADR-0001 Phase 3, Stage 3
 #
 # Deployed by cutover.sh (Phase 7) as an UPDATE to the SAME "postgresqlha"
@@ -41,7 +41,21 @@
 # Once this is live and all consumers have verified (cutover.sh Phase 9),
 # legacy is scaled to 0 (Phase 10) and this becomes the sole answer for
 # the `postgresql`/`db` names on postgresql_db-backend.
-# ──────────────────────────────────────────────────────────────────────────
+#
+# ── primary_slot_name (2026-08-04) ──────────────────────────────────────
+# patroni-0/patroni-1's standby_cluster blocks below include
+# `primary_slot_name: standby_leader_slot`, added to fix a real bootstrap
+# failure (basebackup succeeded but the resulting Postgres process got
+# stuck in "starting" forever because legacy had no replication slot
+# protecting the WAL it needed to resume streaming — recycled by normal
+# checkpoint activity during the 8-11 min basebackup window). Full
+# incident detail and the matching production-side fix (a physical slot
+# named standby_leader_slot created on legacy, plus wal_keep_size bumped
+# to 4GB as defense-in-depth) are documented in postgresql-ha-staging.yaml's
+# header and in the "ADR-0001 Dry-Run Debugging Log" note. This value MUST
+# stay identical to the staging file's per the byte-identical requirement
+# noted above — if you ever change one, change both.
+# ─────────────────────────────────────────────────────────────────────────
 version: "3.6"
 
 services:
@@ -149,6 +163,7 @@ services:
             standby_cluster:
               host: postgresql
               port: 5432
+              primary_slot_name: standby_leader_slot
               create_replica_methods:
               - basebackup_fast_xlog
     secrets:
@@ -198,6 +213,7 @@ services:
             standby_cluster:
               host: postgresql
               port: 5432
+              primary_slot_name: standby_leader_slot
               create_replica_methods:
               - basebackup_fast_xlog
     secrets: