We currently have two ways to generate unique integer IDs:
- timestamp-based: https://docs.pgdog.dev/features/sharding/unique-ids/
- sequence-based (shard-local): https://docs.pgdog.dev/features/sharding/sequences/
We need a 3rd way, which creates global, monotonic, int32 numbers. This will be used with omnisharded tables and on deployments that can't migrate to bigint.
No suggested implementation. There are a few options, all suboptimal: either have a single point of failure or are slow because of required consensus.
We currently have two ways to generate unique integer IDs:
We need a 3rd way, which creates global, monotonic,
int32numbers. This will be used with omnisharded tables and on deployments that can't migrate tobigint.No suggested implementation. There are a few options, all suboptimal: either have a single point of failure or are slow because of required consensus.