Partitioning: Use bigint primary key by default
For new tables, we generally use bigint
for the primary key. Typically, large tables are good candidates for partitioning and large tables require bigint
primary keys at some point.
We already have a couple of examples for tables where integer overflow risk cannot be ignored.
The idea is to address this concern while we partition tables:
- When we create a copy of the table, create the table with a
bigint
primary key. - Install triggers, copy data over etc
- Switch
So this would solve both concerns at the same time: partitioning plus the overflow risk.
Edited by Andreas Brandl