Re: BUG #18027: Logical replication taking forever

Поиск
Список
Период
Сортировка
От Amit Kapila
Тема Re: BUG #18027: Logical replication taking forever
Дата
Msg-id CAA4eK1JMoa3GaNRGTafP2RznW50vV5snVF0+nhRo85uV9PwTkA@mail.gmail.com
обсуждение исходный текст
Ответ на Re: BUG #18027: Logical replication taking forever  (Andres Martin del Campo Campos <andres@invisible.email>)
Ответы Re: BUG #18027: Logical replication taking forever  (Andres Martin del Campo Campos <andres@invisible.email>)
Список pgsql-bugs
On Thu, Jul 20, 2023 at 10:46 PM Andres Martin del Campo Campos <andres@invisible.email> wrote:
Seems like I don't have that table


 image.png


There are no errors in the logs but I only see dead tuples and no live tuples


oh, can you show us the dead and live tuple count on both publisher and subscriber? Ideally, COPY command should only copy the recent data based on the snapshot. It shouldn't copy the old/dead rows. One possibility I could think of is that due to some reason, if there is a failure during the initial sync process, it will ROLLBACK the whole copy and restart it again. So, that way one can see the table is growing with dead tuples and the copy is never finished especially if such an error occurs repeatedly. If that happens, you must see some error in the subscriber-side logs. Can you ensure in some way that such a phenomenon is not happening in your case?

--
With Regards,
Amit Kapila.
Вложения

В списке pgsql-bugs по дате отправления:

Предыдущее
От: Tom Lane
Дата:
Сообщение: Re: BUG #18014: Releasing catcache entries makes schema_to_xmlschema() fail when parallel workers are used
Следующее
От: Alexander Lakhin
Дата:
Сообщение: Re: BUG #18031: Segmentation fault after deadlock within VACUUM's parallel worker