So I’m currently setting up a system in AWS RDS where we are using a postgres database. Rather than connecting to the main database, I was thinking of creating a read replica to use as the source database for cubejs. However it seems that the following things are true:
- No one can write to a postgres read replica under any circumstances
- By default, CubeJS wishes to write to the source database (which is currently the read replica)
- We can turn the readOnly option on, but we are working with large production workloads and the warning states
“Read only pre-aggregations are only suitable for small datasets since they require loading all the data into Cube.js process memory. We do not recommend using readOnly mode for production workloads.”
It seems that the only way is instead of using the RDS read replica is to use AWS DMS to create a full replica of the database and then use that as the source db. Is that correct? Or is there a simpler way that someone has figured out to do this? Would it be better to just hook it up to our main db or would the throughput be more likely to crash everything? Or just turn readOnly on?
Any and all help is appreciated. Thank you!