Thread

Re: Sequence Access Methods, round two

Andrei Lepikhov <lepihov@gmail.com> — 2025-11-07T16:05:55Z

On 7/11/2025 07:12, Michael Paquier wrote:
> On Thu, Oct 30, 2025 at 04:44:23PM +0900, Michael Paquier wrote:
>> Rebased due to conflicts with 76acf4b722fa.
> 
> Rebased as v24 as GUCs need to be in alphabetical order.

Thanks for your efforts!
I haven't dived into the code yet; I just want to add some words about 
the reasoning.
1. For pgEdge Snowflake, it would be really beneficial - there are not 
only code copying issues, but also the Serial -> Snowflake conversion 
algorithm. Additionally, for now, only an in-core sequence can be used 
as an 'identity always' column.
2. Multimaster project [1] also attempts to resolve the sequence issue. 
You may check how its developers tackle the problem - I guess any 
active-active configuration needs it.
3. I wonder if postgres_fdw-based shardman could benefit from that. I 
imagine an implementation of an auto-partitioned table where the 
partitioning key is generated automatically based on the partition where 
the tuple is inserted.
4. Any kind of non-uniform distribution may be implemented that is 
helpful for testing purposes

[1] https://github.com/postgrespro/mmts

-- 
regards, Andrei Lepikhov,
pgEdge