fix(ods): изменен батчевый резолвер для транзакционных таблиц

- Зачем:
  - текущая реализация ODS batch_id теряла данные транзакционных таблиц, если между запусками ODS STG успевал отработать дважды (брался только последний батч).
- Что:
  - изменены скрипты загрузки транзакционных таблиц (bookings, tickets, flights, segments, boarding_passes) для использования паттерна HWM по _load_ts вместо фильтрации по конкретному батчу.
  - обновлен комментарий в DAG bookings_to_gp_ods, объясняющий разное поведение для справочников и транзакционных данных.
  - сохранено использование оригинального batch_id из STG для поля _load_id в слое ODS для сквозного трассирования.
- Проверка:
  - запуск пайплайнов и проверка, что все батчи загружаются из STG в ODS без потерь.
This commit is contained in:
2026-03-01 18:15:04 +03:00
parent faf410b4e8
commit cfd20328d6
6 changed files with 29 additions and 18 deletions
+9
View File
@@ -34,6 +34,15 @@ def _resolve_stg_batch_id(**context) -> str:
"""
Возвращает stg_batch_id из dag_run.conf или вычисляет последний согласованный батч.
ВНИМАНИЕ: Это значение используется ТОЛЬКО для загрузки snapshot-справочников
(airports, airplanes, routes, seats).
Для инкрементальных транзакционных таблиц (bookings, tickets, flights, segments,
boarding_passes) этот батч НЕ используется. Вместо этого они грузят все новые
записи по HWM: WHERE load_dttm > (SELECT MAX(_load_ts) FROM ods.table).
Это сделано для того, чтобы не потерять инкременты, если STG-DAG запускался
несколько раз до запуска ODS-DAG'а.
Зачем нужна согласованность (INTERSECT по всем справочникам)?
Чтобы ODS загружал только те данные, для которых уже приехали ВСЕ связанные
справочники. Это защищает от рассинхрона данных, когда часть измерений в текущем
+3 -3
View File
@@ -14,7 +14,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.boarding_passes AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.boarding_passes)
)
UPDATE ods.boarding_passes AS o
SET seat_no = s.seat_no,
@@ -48,7 +48,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.boarding_passes AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.boarding_passes)
)
INSERT INTO ods.boarding_passes (
ticket_no,
@@ -67,7 +67,7 @@ SELECT
s.boarding_no,
s.boarding_time,
s.event_ts,
'{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
s.batch_id,
now()
FROM src AS s
WHERE s.rn = 1
+6 -4
View File
@@ -7,18 +7,19 @@ WITH src AS (
NULLIF(s.book_date, '')::TIMESTAMP WITH TIME ZONE AS book_date,
NULLIF(s.total_amount, '')::NUMERIC(10,2) AS total_amount,
s.src_created_at_ts AS event_ts,
s.batch_id,
ROW_NUMBER() OVER (
PARTITION BY s.book_ref
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.bookings AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.bookings)
)
UPDATE ods.bookings AS o
SET book_date = s.book_date,
total_amount = s.total_amount,
event_ts = s.event_ts,
_load_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
_load_id = s.batch_id,
_load_ts = now()
FROM src AS s
WHERE s.rn = 1
@@ -36,12 +37,13 @@ WITH src AS (
NULLIF(s.book_date, '')::TIMESTAMP WITH TIME ZONE AS book_date,
NULLIF(s.total_amount, '')::NUMERIC(10,2) AS total_amount,
s.src_created_at_ts AS event_ts,
s.batch_id,
ROW_NUMBER() OVER (
PARTITION BY s.book_ref
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.bookings AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.bookings)
)
INSERT INTO ods.bookings (
book_ref,
@@ -56,7 +58,7 @@ SELECT
s.book_date,
s.total_amount,
s.event_ts,
'{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
s.batch_id,
now()
FROM src AS s
WHERE s.rn = 1
+5 -5
View File
@@ -7,7 +7,7 @@ WITH segment_flights AS (
SELECT DISTINCT
NULLIF(s.flight_id, '')::INTEGER AS flight_id
FROM stg.segments AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.segments)
AND s.flight_id IS NOT NULL
AND s.flight_id <> ''
),
@@ -40,7 +40,7 @@ src_union AS (
f.event_ts,
f.load_dttm
FROM stg_flights_typed AS f
WHERE f.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE f.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.flights)
UNION ALL
@@ -103,7 +103,7 @@ WITH segment_flights AS (
SELECT DISTINCT
NULLIF(s.flight_id, '')::INTEGER AS flight_id
FROM stg.segments AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.segments)
AND s.flight_id IS NOT NULL
AND s.flight_id <> ''
),
@@ -135,7 +135,7 @@ src_union AS (
f.event_ts,
f.load_dttm
FROM stg_flights_typed AS f
WHERE f.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE f.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.flights)
UNION ALL
@@ -190,7 +190,7 @@ SELECT
s.actual_departure,
s.actual_arrival,
s.event_ts,
'{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
s.batch_id,
now()
FROM src AS s
WHERE s.rn = 1
+3 -3
View File
@@ -13,7 +13,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.segments AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.segments)
)
UPDATE ods.segments AS o
SET fare_conditions = s.fare_conditions,
@@ -44,7 +44,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.segments AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.segments)
)
INSERT INTO ods.segments (
ticket_no,
@@ -61,7 +61,7 @@ SELECT
s.fare_conditions,
s.segment_amount,
s.event_ts,
'{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
s.batch_id,
now()
FROM src AS s
WHERE s.rn = 1
+3 -3
View File
@@ -14,7 +14,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.tickets AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.tickets)
)
UPDATE ods.tickets AS o
SET book_ref = s.book_ref,
@@ -49,7 +49,7 @@ WITH src AS (
ORDER BY s.src_created_at_ts DESC NULLS LAST, s.load_dttm DESC
) AS rn
FROM stg.tickets AS s
WHERE s.batch_id = '{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text
WHERE s.load_dttm > (SELECT COALESCE(MAX(_load_ts), '1900-01-01 00:00:00'::TIMESTAMP) FROM ods.tickets)
)
INSERT INTO ods.tickets (
ticket_no,
@@ -68,7 +68,7 @@ SELECT
s.passenger_name,
s.is_outbound,
s.event_ts,
'{{ ti.xcom_pull(task_ids="resolve_stg_batch_id") }}'::text,
s.batch_id,
now()
FROM src AS s
WHERE s.rn = 1