fix(mnesia): per-slot volumes hint and harden verification lookup.

Avoid shared Mnesia dir across replicas; make get_or_create_token and CT verify_user resilient to brief cross-node lag.

Co-authored-by: Cursor <cursoragent@cursor.com>

docker/docker-compose.swarm.yml

@@ -76,12 +76,17 @@ services:
           - eventhub
           - eventhub-node
     volumes:
+      # Per-task volume when scaling replicas > 1 (never share one Mnesia dir).
       - type: volume
-        source: eventhub-data
+        source: eventhub-data-{{.Task.Slot}}
         target: /app/data
     deploy:
       replicas: 1
       endpoint_mode: dnsrr
+      update_config:
+        parallelism: 1
+        delay: 30s
+        order: stop-first
       restart_policy:
         condition: any
       labels:
@@ -222,8 +227,8 @@ networks:
     driver: overlay
 
 volumes:
-  eventhub-data:
-#    name: 'eventhub-data-{{.Task.Slot}}'
+  # eventhub-data-{{.Task.Slot}} created by Swarm from service mount template
+  # (do not use a single shared volume for multi-replica Mnesia)
   prometheus-data:
   grafana-data:
   traefik-logs:

src/core/core_verification.erl

@@ -34,17 +34,19 @@ verify_token(Token) ->
 
 %%%-------------------------------------------------------------------
 %%% @doc Возвращает существующий токен пользователя или создаёт новый.
+%%% Допускает несколько токенов на user_id (register + повторные вызовы).
 %%% @end
 %%%-------------------------------------------------------------------
 -spec get_or_create_token(UserId :: binary()) ->
   {ok, Token :: binary(), ExpiresAt :: calendar:datetime()} | {error, not_found}.
 get_or_create_token(UserId) ->
-  case mnesia:dirty_match_object(#verification{user_id = UserId, _ = '_'}) of
-    [V] -> {ok, V#verification.token, V#verification.expires_at};
-    [] ->
-      case core_user:get_by_id(UserId) of
-        {ok, _} -> create_token(UserId);
-        Error -> Error
+  case find_token(UserId) of
+    {ok, Token, ExpiresAt} ->
+      {ok, Token, ExpiresAt};
+    not_found ->
+      case user_exists(UserId) of
+        true -> create_token(UserId);
+        false -> {error, not_found}
       end
   end.
 
@@ -56,3 +58,23 @@ get_or_create_token(UserId) ->
 delete_token(Token) ->
   mnesia:dirty_delete(verification, Token),
   ok.
+
+%%--------------------------------------------------------------------
+find_token(UserId) ->
+  case mnesia:dirty_match_object(#verification{user_id = UserId, _ = '_'}) of
+    [#verification{token = Token, expires_at = ExpiresAt} | _] ->
+      {ok, Token, ExpiresAt};
+    [] ->
+      not_found
+  end.
+
+%% dirty_read first; transactional read as fallback for cluster replica lag.
+user_exists(UserId) ->
+  case core_user:get_by_id(UserId) of
+    {ok, _} -> true;
+    {error, not_found} ->
+      case mnesia:transaction(fun() -> mnesia:read(user, UserId) end) of
+        {atomic, [_ | _]} -> true;
+        _ -> false
+      end
+  end.

src/infra/infra_mnesia.erl

@@ -229,7 +229,7 @@ add_local_ram_copy(Tab) ->
   end.
 
 wait_for_tables_available() ->
-  lists:foreach(fun(Tab) -> wait_for_table(Tab) end, ?DISC_TABLES).
+  lists:foreach(fun(Tab) -> wait_for_table(Tab) end, ?DISC_TABLES ++ ?RAM_TABLES).
 
 wait_for_table(Tab) ->
   case lists:member(Tab, mnesia:system_info(tables)) of
@@ -260,6 +260,12 @@ prune_dead_nodes() ->
         false -> ok
       end
                   end, ?DISC_TABLES),
+    lists:foreach(fun(Tab) ->
+      case lists:member(Node, mnesia:table_info(Tab, ram_copies)) of
+        true -> catch mnesia:del_table_copy(Tab, Node);
+        false -> ok
+      end
+                  end, ?RAM_TABLES),
     catch mnesia:del_table_copy(schema, Node)
                 end, DeadNodes).
 
@@ -327,6 +333,7 @@ create_indices() ->
   mnesia:add_table_index(calendar_specialist, user_id),
   mnesia:add_table_index(user, nickname),
   mnesia:add_table_index(user, email),
+  mnesia:add_table_index(verification, user_id),
   mnesia:add_table_index(notification, user_id),
   mnesia:add_table_index(notification, is_read),
   mnesia:add_table_index(auth_session, family_id),

test/api/api_test_runner.erl

@@ -402,13 +402,25 @@ create_event(Token, CalId, Params) ->
 
 %% @doc Подтверждает email пользователя, используя админский эндпоинт
 %% для получения верификационного токена и публичный /v1/verify.
+%% Retries briefly: IFT Traefik may hit another replica before Mnesia settles.
 -spec verify_user(AdminToken :: binary(), UserId :: binary()) -> ok.
 verify_user(AdminToken, UserId) ->
-  #{<<"token">> := Token} = admin_get(
-    <<"/v1/admin/users/", UserId/binary, "/verification-token">>, AdminToken),
-  {ok, 200, _, _} = client_request(post, <<"/v1/verify">>, <<>>,
-    jsx:encode(#{<<"token">> => Token})),
-  ok.
+  verify_user(AdminToken, UserId, 8).
+
+verify_user(AdminToken, UserId, RetriesLeft) ->
+  Path = <<"/v1/admin/users/", UserId/binary, "/verification-token">>,
+  case admin_request(get, Path, AdminToken) of
+    {ok, 200, _, Body} ->
+      #{<<"token">> := Token} = jsx:decode(list_to_binary(Body), [return_maps]),
+      {ok, 200, _, _} = client_request(post, <<"/v1/verify">>, <<>>,
+        jsx:encode(#{<<"token">> => Token})),
+      ok;
+    {ok, 404, _, _} when RetriesLeft > 1 ->
+      timer:sleep(100),
+      verify_user(AdminToken, UserId, RetriesLeft - 1);
+    Other ->
+      error({verify_user_failed, UserId, Other})
+  end.
 
 %%%===================================================================
 %%% Внутренние функции