Hello do I understand correctly from the documentation that you can not change how the table is partitioned? For example if I partitioned a table by day and tenant, but now I realize it should be partitioned by month and tenant there is nothing I can do? In druid for example the partition strategy is tied not to the table but to the ingestion process. So I can change it and repartition the existing partitions atomically with REPLACE. In starrocks I see no other way than to create a new table, temporarily load to new and old table at the same time, copy old data over to the new table and then SWAP them. Is there a better way?

https://docs.starrocks.io/docs/table_design/data_distribution/expression_partitioning/#merge-expression-partitions Could you try it? This is available since v3.5

I saw that but that only alters the existing partitions, not change the strategy for new partitions that will be created in the future, right?

Yeah, you are correct. This is mainly used for archiving. For now, creating a new table and swapping is the right way to handle it. But I think this requirement makes sense. Would you like to create a ticket on GitHub for us to track this requirement?

No I don't really need it right now. We don't have something productive yet. I am working on a proof of concept while exploring technology options. Just wanted to understand what is possible, since I cannot learn to swim after I fall in the water.

Got it, makes sense

But for sure would be good feature if the underlying architecture of the database allows for it without great cost.