On Mon, Apr 15, 2024 at 10:12 PM David Steele <da...@pgmasters.net> wrote: > Anyway, I think it should be fixed or documented as a caveat since it > causes a hard failure on restore.
Alright, I'll look into this. > I know Tomas added some optimizations that work best with --no-manifest > but if we can eventually read compressed tars (which I expect to be the > general case) then those optimizations are not very useful. My belief is that those optimizations work fine with or without manifests; you only start to lose the benefit in cases where you use different checksum types for different backups that you then try to combine. Which should hopefully be a rare case. -- Robert Haas EDB: http://www.enterprisedb.com