Thread: RUM and WAL-generation

Hi all.

 

The RUM-index is very nice, but indexes get very large and produce insane amounts of WAL.

Due to the way PG works (IIUC), updating an int-column in a row produces lots of WAL because the whole row is duplicated, and if that row holds RUM-indexed columns it gets really bad...

We hav actually run out of WAL-space in production because of this.

 

I see this TODO-entry in RUM:

 

What is the status on this?