On Tue, 2026-09-15 at 07:08 +0530, Brajendra Pratap Singh wrote:
> I have performed the production database upgrade from postgresql 13 to 17 version
> using pg_upgrade --link method on ubuntu 22.04 successfully.
>
> Now after few days of upgrade, while dropped one database of size 900+ GB
> successfully, the space not get reclaimed physically from mount point /usr/pgdata.
>
> Old PG13 data directory: /usr/pgdata/data1 --> size 4.5TB
>
> New PG17 data directory:
> /usr/pgdata/data17 --> 5.6GB
>
> I didn't perform the delete_old_cluster.sh yet.
>
> I can see that the oid of this dropped database is present in PG13 data directory
> base location and occupying 972GB size, while it's not present in PG17 data
> directory base location.
>
> Now my doubt is if I removed Old data directory, will it cause any impact on
> running PG17 cluster.
>
> Please suggest the best and safe approach to reclaim the space.
Just go ahead and remove the old cluster.
With pg_upgrade --link, the data files exist only once, but are linked
in the data directory on both the old and the new server.
Try
du -sm /path/to/old/cluster /path/to/new/cluster
and the old cluster will appear large, while the new cluster will seem small
But if you run
du -sm /path/to/new/cluster /path/to/old/cluster
the new cluster will look big, and the new one will look small.
That is because the same files are referenced in both directories.
Yours,
Laurenz Albe
