[grzm]

This can indeed be an issue, with both indexes and tables. For tables, I've had success with tuning fillfactor to take into account update patterns. If there's a deterministic subset of rows that are updated more frequently, splitting those into a smaller table/partition, perhaps with a different fillfactor can help as well. This also makes vacuum faster for this subset. Even if there isn't a deterministic subset, partitioning for smaller table size can speed up vacuum regardless.

For indexes, using concurrent reindex to create new indexes and dropping the old can make a difference.

Yeah, it would be nice if more of this were automated and it's good that there's work being done. There are some options available in existing that can ameliorate bloat.

[jacobkg]

Solid advice! I remember reading about fillfactor but didn't get a chance to try it out on our workload.