add security test 03

alembic/versions/0032_repair_missing_email_columns.py

@@ -0,0 +1,57 @@
+"""repair missing email auth columns in legacy prod databases
+
+Revision ID: 0032_email_cols_fix
+Revises: 0031_pii_retention_and_consent
+Create Date: 2026-03-02
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "0032_email_cols_fix"
+down_revision = "0031_pii_retention_and_consent"
+branch_labels = None
+depends_on = None
+
+
+def _has_column(inspector: sa.Inspector, table: str, column: str) -> bool:
+    return any(str(col.get("name")) == column for col in inspector.get_columns(table))
+
+
+def _has_index(inspector: sa.Inspector, table: str, index_name: str) -> bool:
+    return any(str(idx.get("name")) == index_name for idx in inspector.get_indexes(table))
+
+
+def upgrade() -> None:
+    bind = op.get_bind()
+    inspector = sa.inspect(bind)
+
+    if not _has_column(inspector, "clients", "email"):
+        op.add_column("clients", sa.Column("email", sa.String(length=255), nullable=True))
+    inspector = sa.inspect(bind)
+    if not _has_index(inspector, "clients", "ix_clients_email"):
+        op.create_index("ix_clients_email", "clients", ["email"], unique=False)
+
+    if not _has_column(inspector, "requests", "client_email"):
+        op.add_column("requests", sa.Column("client_email", sa.String(length=255), nullable=True))
+    inspector = sa.inspect(bind)
+    if not _has_index(inspector, "requests", "ix_requests_client_email"):
+        op.create_index("ix_requests_client_email", "requests", ["client_email"], unique=False)
+
+    if not _has_column(inspector, "otp_sessions", "channel"):
+        op.add_column("otp_sessions", sa.Column("channel", sa.String(length=16), nullable=True, server_default="SMS"))
+    if not _has_column(inspector, "otp_sessions", "email"):
+        op.add_column("otp_sessions", sa.Column("email", sa.String(length=255), nullable=True))
+    op.execute("UPDATE otp_sessions SET channel = 'SMS' WHERE channel IS NULL")
+    op.alter_column("otp_sessions", "channel", nullable=False, server_default=None)
+    inspector = sa.inspect(bind)
+    if not _has_index(inspector, "otp_sessions", "ix_otp_sessions_email"):
+        op.create_index("ix_otp_sessions_email", "otp_sessions", ["email"], unique=False)
+
+
+def downgrade() -> None:
+    # Intentionally no-op for safety on heterogeneous legacy databases.
+    pass

context/11_test_runbook.md

@@ -206,3 +206,4 @@ echo $?  # 0=OK, >0=ALERT
 - `make security-smoke` — `PASS`, отчет: `reports/security/security-smoke-20260302-125811.md`.
 - `./scripts/ops/security_smoke.sh https://ruakb.online` — `PASS`, отчет: `reports/security/security-smoke-20260302-130511.md` (TLS и security headers внешнего контура).
 - `./scripts/ops/security_smoke.sh https://ruakb.ru` — `PASS`, отчет: `reports/security/security-smoke-20260302-130536.md` (TLS и security headers внешнего контура).
+- `docker compose run --rm backend alembic upgrade head` — успешно после hotfix-миграции `0032_email_cols_fix` (ремонт legacy-схемы: `requests.client_email`, `otp_sessions.channel/email`, индексы).

scripts/ops/prod_security_audit.sh

@@ -78,8 +78,20 @@ ensure_compose_files() {
 }
 
 stack_up_and_migrate() {
-  log "Starting production stack (nginx profile)"
-  "${PROD_COMPOSE[@]}" up -d --build --remove-orphans db redis minio backend chat-service email-service worker beat frontend edge clamav
+  log "Starting core infra services"
+  "${PROD_COMPOSE[@]}" up -d --build --remove-orphans --force-recreate db redis minio clamav
+
+  log "Starting app services (backend/chat/email/worker/beat)"
+  "${PROD_COMPOSE[@]}" up -d --build --remove-orphans --force-recreate backend chat-service email-service worker beat
+
+  log "Waiting app services to become healthy"
+  wait_service_healthy "backend" 60
+  wait_service_healthy "chat-service" 60
+  wait_service_healthy "email-service" 60
+
+  # Force recreate frontend/edge after chat/backend to avoid stale DNS upstream cache in nginx.
+  log "Starting/recreating frontend and edge"
+  "${PROD_COMPOSE[@]}" up -d --build --remove-orphans --force-recreate frontend edge
 
   log "Applying migrations"
   "${PROD_COMPOSE[@]}" exec -T backend alembic upgrade head
@@ -94,6 +106,30 @@ print("production security config validation: ok")
 PY
 }
 
+wait_service_healthy() {
+  local service="$1"
+  local max_attempts="${2:-60}"
+  local attempt=1
+  local cid=""
+  local status=""
+
+  while (( attempt <= max_attempts )); do
+    cid="$("${PROD_COMPOSE[@]}" ps -q "$service" 2>/dev/null || true)"
+    if [[ -n "$cid" ]]; then
+      status="$(docker inspect --format '{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}' "$cid" 2>/dev/null || true)"
+      case "$status" in
+        healthy|running)
+          return 0
+          ;;
+      esac
+    fi
+    sleep 2
+    attempt=$((attempt + 1))
+  done
+
+  fail "Service '${service}' did not become healthy in time (last_status=${status:-unknown})"
+}
+
 run_local_smoke() {
   if [[ "$SKIP_LOCAL_SMOKE" == "1" ]]; then
     log "Skipping local smoke checks (SKIP_LOCAL_SMOKE=1)"