[Spidler]

But you cannot cluster on a BRIN index, and isn't cluster an operation that you need to redo regularly?

I thought so, since while the cluster flag is set on an index, it won't guarantee future writes to be clustered.

Since you can issue the cluster command to update the clustering.

So, I wonder how you are to maintain the setup.

[cldellow]

As others have mentioned, if your workload is mostly appends and the data is appended in the correct order, you don't need to redo the clustering. In our case, the clustering was a one-off to get the data in the right shape.

(Although if you do need to redo clustering, look into pg_repack.)