Marc dirix wrote: >Can't this be solved with partitioning the bodyparts table?
That would probably ease the pain with regard to vacuuming of the smaller bodyparts, but it would not solve the problem with a database restore. Restoring a 413GB database is impossible. Not if you want to keep the downtime limited to something like less than one hour. People need their email. They're not going to like it when you tell them that due a restore in progress, all mail will be unavailable for the next 48 hours. Also, a vacuum on the larger bodyparts still is a disaster, since it involves way too much copying of data. Postgres stores data in 1GB chunks. Did you know that if a vacuum is interrupted halfway through such a file, it will restart from the beginning of that file again? -- Stephen. This is a day for firm decisions! Or is it?
