fix(deploy): migrate 跨 pod 串行化 + rescale 迁移幂等守卫(测试环境×10⁶事故复盘)

事故:连推 3 commit 触发 3 轮 rolling 部署,多个新 pod 并发跑 migrate 且崩溃重跑,
积分 ×10 rescale 被交错重放——ai.0025 执行 6 次(单价/任务计价 ×10⁶),accounts.0008
执行 2 次(限额 ×100),billing.0003/0004 从未完成(django_migrations 漏记录)。
测试库数据已按精确倍率手工修复并补记迁移记录(备份于本机)。

两层防复发:
1. docker-entrypoint 用 MySQL GET_LOCK('airshelf_migrate') 串行化 migrate,
   后到 pod 等锁,拿到时迁移已被记录 → 自然 no-op;
2. 三个 rescale 迁移加 airshelf_rescale_marker 幂等标记(与数据变更同事务提交):
   记录丢失/崩溃重跑时,标记在 → 跳过,不会重复 ×10。

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

core/backend/apps/ai/migrations/0025_points_rescale.py

@@ -20,10 +20,41 @@ REVERSE_STATEMENTS = [
 ]
 
 
+# 幂等守卫(2026-07-07 测试环境事故复盘):并发/崩溃后的 migrate 重跑会把 ×10 重复应用
+# (事故中本迁移被交错重放,单价被乘到 ×10⁶)。标记行与数据变更同一事务提交:
+# 上次成功 → 标记在 → 跳过;上次崩溃回滚 → 标记不在 → 安全重放。
+MARKER = "ai.0025_points_rescale"
+_MARKER_DDL = "CREATE TABLE IF NOT EXISTS airshelf_rescale_marker (name varchar(80) NOT NULL PRIMARY KEY)"
+
+
 def _run(statements):
     def apply(apps, schema_editor):
-        with transaction.atomic(using=schema_editor.connection.alias):
-            with schema_editor.connection.cursor() as cursor:
+        conn = schema_editor.connection
+        with conn.cursor() as cursor:  # DDL 幂等,MySQL 隐式提交故放事务外
+            cursor.execute(_MARKER_DDL)
+        with transaction.atomic(using=conn.alias):
+            with conn.cursor() as cursor:
+                cursor.execute("SELECT COUNT(*) FROM airshelf_rescale_marker WHERE name = %s", [MARKER])
+                if cursor.fetchone()[0]:
+                    return  # 已应用过(django_migrations 记录丢失/并发重跑),幂等跳过
+                cursor.execute("INSERT INTO airshelf_rescale_marker (name) VALUES (%s)", [MARKER])
+                for sql in statements:
+                    cursor.execute(sql)
+
+    return apply
+
+
+def _run_reverse(statements):
+    def apply(apps, schema_editor):
+        conn = schema_editor.connection
+        with conn.cursor() as cursor:
+            cursor.execute(_MARKER_DDL)
+        with transaction.atomic(using=conn.alias):
+            with conn.cursor() as cursor:
+                cursor.execute("SELECT COUNT(*) FROM airshelf_rescale_marker WHERE name = %s", [MARKER])
+                if not cursor.fetchone()[0]:
+                    return  # 未应用过或已回滚,无需反向
+                cursor.execute("DELETE FROM airshelf_rescale_marker WHERE name = %s", [MARKER])
                 for sql in statements:
                     cursor.execute(sql)
 
@@ -32,4 +63,4 @@ def _run(statements):
 
 class Migration(migrations.Migration):
     dependencies = [("ai", "0024_aitask_base_cost")]
-    operations = [migrations.RunPython(_run(RESCALE_STATEMENTS), _run(REVERSE_STATEMENTS))]
+    operations = [migrations.RunPython(_run(RESCALE_STATEMENTS), _run_reverse(REVERSE_STATEMENTS))]