On Tue, Jun 26, 2018 at 8:38 AM Alexander Shutyaev <shutyaev@xxxxxxxxx> wrote: > > Hello again, > > I've performed another test - I've migrated to the new cluster using dump restore: pg_dumpall | psql. It went well, although it took 6 days while pg_upgrade usually took a night. > > Is there any hope the issue with pg_upgrade can be resolved? If not, could you give me some hints as to how can I decrease time needed for pg_dumpall | psql? If you use pg_dump with the custom format you can use pg_restore --jobs=n to do a parallel restore. I think you can do pg_dumpall --globals-only to restore users, and then do a pg_dump per database. Groeten, Arjen
