As usual, you are totally right Tom. But thanks everybody for the debate. Only I have to add, is near a "must" to run with --check before actually run the command.
2013/5/16 Tom Lane <t...@sss.pgh.pa.us> > Fabio Rueda Carrascosa <avances...@gmail.com> writes: > > Hello, I have a 9.1 cluster with 50 databases, only one table per db with > > 2000 rows only, but a lot of schema around each one (postgis databases) > > > The old cluster size is 1GB > > > du -chs /var/lib/postgresql/9.1/main/ > > 1.1G > > > now I run a pg_upgrade to 9.2 with hard link mode, > > > pg_upgrade --link \ > > --old-datadir=/var/lib/postgresql/9.1/main \ > > --new-datadir=/var/lib/postgresql/9.2/main \ > > --old-bindir=/usr/lib/postgresql/9.1/bin \ > > --new-bindir=/usr/lib/postgresql/9.2/bin > > > du -chs /var/lib/postgresql/9.2/main/ > > 880M > > > Is the expected behaviour? I can't double the space in production. > > I don't think anybody actually answered your original question. > The above doesn't represent a doubling of disk space, it just shows that > "du" only tells you how much file space is linked into the directory > tree you ask it about. That is, there's lots of overlap between the > first and second du results. If you try "du" passing it both directory > trees together, it should give you a number for the total space > consumption that's not much more than 1.1G. (Depending on which version > of "du" you're using, you may need to give it an additional switch to > tell it not to double-count multiply-linked files.) > > regards, tom lane >
