Hi,
Alsready tried to use --link and --jobs, but you cannot ommit the
"select lo_unlink ...." for every rows containing datatype text in
your database that the pg_* program creates in the export/dump file.
These select's takes a lot of time, hours ...
The pg_dumpall => size of the dumpfile is much more than the size of
exported database
Kind regards
On 20/07/2023 15:29, Edward J. Sabol wrote:
On Jul 20, 2023, at 7:23 AM, Jef Mortelle <jefmortelle@xxxxxxxxx> wrote:
I use /usr/lib/postgresql15/bin/pg_upgrade -v -p 5431 -P 5432
I takes a lot of time, more than 6 hours
I'm in the middle of a similar database migration myself, but I'm coming from 10.23.
Have you considered adding the --link and --jobs=NN (where NN is the number of CPU cores on your server minus 2 or something like that) to your pg_upgrade command? I wonder if that would speed things up significantly.
Also, if your file system supports reflinks (Linux kernel 4.5 and XFS with reflink support activated or Btrfs or macOS APFS), then use --clone instead of --link.
Good luck,
Ed
