fix: subscribe stats_collector after Mnesia tables are ready
CI / test (push) Failing after 13m10s
CI / deploy-ift (push) Has been skipped
CI / e2e-ift (push) Has been skipped
CI / deploy-stage (push) Has been skipped
CI / e2e-stage (push) Has been skipped

Backfill ran during init_tables before wait_for_tables, so dirty reads hit {no_exists,user} and crash-looped eventhub on IFT.
This commit is contained in:
2026-07-18 23:33:56 +03:00
parent fa07d91f4b
commit 31c2a1b7d7
4 changed files with 28 additions and 4 deletions
+2
View File
@@ -52,6 +52,8 @@ start_application() ->
ok = infra_mnesia:init_tables(),
ok = infra_mnesia:wait_for_tables(),
ok = migration_engine:ensure_applied(),
%% После wait + migrations: иначе backfill ловит {no_exists, user} на ещё не загруженных таблицах.
ok = stats_collector:subscribe(),
calendar_html_renderer:init_cache(),
application:ensure_all_started(cowboy),
start_http(), % Пользовательский API (8080)