RE: Multi-Master Logical Replication

Поиск
Список
Период
Сортировка
От kuroda.hayato@fujitsu.com
Тема RE: Multi-Master Logical Replication
Дата
Msg-id TYAPR01MB5866967CC3C287AD6D3FFE53F5D09@TYAPR01MB5866.jpnprd01.prod.outlook.com
обсуждение исходный текст
Ответ на Re: Multi-Master Logical Replication  (Amit Kapila <amit.kapila16@gmail.com>)
Ответы RE: Multi-Master Logical Replication  ("kuroda.hayato@fujitsu.com" <kuroda.hayato@fujitsu.com>)
Список pgsql-hackers
Hi hackers,

I created a small PoC. Please see the attached patches.

REQUIREMENT

Before patching them, patches in [1] must also be applied.


DIFFERENCES FROM PREVIOUS DESCRIPTIONS

* LRG is now implemented as SQL functions, not as a contrib module.
* New tables are added as system catalogs. Therefore, added tables have oid column.
* The node_id is the strcat of system identifier and dbid.


HOW TO USE

In the document patch, a subsection 'Example' was added for understanding LRG. In short, we can do

1. lrg_create on one node
2. lrg_node_attach on another node

Also attached is a test script that constructs a three-nodes system.


LIMITATIONS

This feature is under development, so there are many limitations for use case.

* The function for detaching a node from a group is not implemented.
* The function for removing a group is not implemented.
* LRG does not lock system catalogs and databases. Concurrent operations may cause inconsistent state.
* LRG does not wait until the upstream node reaches the latest lsn of the remaining nodes.
* LRG does not support initial data sync. That is, it can work well only when all nodes do not have initial data.


[1]: https://commitfest.postgresql.org/38/3610/

Best Regards,
Hayato Kuroda
FUJITSU LIMITED


Вложения

В списке pgsql-hackers по дате отправления:

Предыдущее
От: Kyotaro Horiguchi
Дата:
Сообщение: Re: create_help.pl treats as replaceable
Следующее
От: Amit Langote
Дата:
Сообщение: Re: First draft of the PG 15 release notes