[Question] Adding new shards to an existing cluster (rebalancing)

[vitabaks]

Hi! I'm continuing to explore PgDog and so far the documentation makes it clear how to shard a monolithic database. However, it’s not entirely clear how to add new shards to an already sharded cluster and then rebalance the data across shards.

Could you please clarify how this workflow is expected to work? If possible, it would also be great to have this scenario described in the documentation.

Thanks for PgDog — it's a really impressive project!

[levkk]

The docs you linked describe the process :)

Resharding changes the number of shards in an existing database cluster in order to add or remove capacity. To make this less impactful on production operations, PgDog's strategy for resharding is to create a new database cluster and reshard data while moving it to the new databases.

Resharding from any number of shards, including 1, currently requires rewriting data to a new cluster. This is mostly because the hashing function we use is uniform so data will have to be moved pretty much everywhere if the number of shards changes. For list-based and range-based, we can definitely reshard in place, but not currently supported. We're going to add rendezvous hashing at some point, and once that's there, we'll start moving data between existing shards.

[vitabaks]

Everything is clear, thanks for the reply.

For very large datasets (hundreds of terabytes), this approach can be quite expensive.
Perhaps it would make sense to add a new shard in advance to slow down the growth of existing shards, since some of the new rows would start being written to the new shard.

P.S. Please consider documenting this scenario explicitly in the documentation.

[levkk]

It definitely would be. We just need a sharding function that works well with that approach. If you're using list-based or range-based sharding, you can do this already, as long as the new rows have different IDs - you won't move existing data but new data will land on new shards.