Why Deleting Data From Delta Tables Makes Them Bigger
Deletion vectors, tombstones, and the VACUUM step that frees storage
A right-to-be-forgotten batch lands in the data team’s queue: erase 12,000 users from a 2TB events table. The engineer on rotation writes the DELETE, the job finishes green, and a count on those user IDs returns zero. Ticket closed, compliance notified, everyone moves on.
Three weeks later the storage dashboard flags the same table. The directory behind …
