Re: PG Sharding
| От | Matej |
|---|---|
| Тема | Re: PG Sharding |
| Дата | |
| Msg-id | CAJB+8mbjN6jSnjj84Qgnzuda7r5j12_9S56zCHybcumz_rNKTQ@mail.gmail.com обсуждение исходный текст |
| Ответ на | Re: PG Sharding (Thomas Boussekey <thomas.boussekey@gmail.com>) |
| Список | pgsql-general |
Thanks Thomas.
Still fancying the manual approach litlle bit more.
Will probably go with 8 database and 32 schema per machine. This way, will keep in limits for administration tools as well as autovacuum, also will be ready for connection pooling, as 8 databases is not to much.
This will give us 256 shard per machine, but will be tunable. The lower number will also prevent to much memory/disk fragmentation and with this bad cache hit ratios.
Will also use monthly partitioning per shard, to reduce the change of big tables forming.
Thanks everyone.
2018-01-30 15:26 GMT+01:00 Thomas Boussekey <thomas.boussekey@gmail.com>:
Using citusdb enterprise, you can replicate the table shards.Here is the link to the documentation:
https://docs.citusdata.com/en/v7.2/reference/user_defined_ functions.html#replicate- table-shards Regards,Thomas2018-01-30 12:18 GMT+01:00 Matej <gmatej@gmail.com>:As already said. It's missing 2 level sharding and is restricted with SPOF.BRMatej2018-01-30 12:05 GMT+01:00 Rakesh Kumar <rakeshkumar464@aol.com>:>We are looking for multi tenancy but at scale. That's why the sharding and partitioning. It depends how you look at the distributed part.Citusdb.
В списке pgsql-general по дате отправления: