On 9/4/24 03:48, Sam Son wrote: > Hi Adrian, Muhammad, > > Thanks for the quick response. > > For new I cannot do changes in old version DB, since it is deployed > remotely and i dont have any access. And it has to be done from multiple > servers. > > As a work around I tried two solutions.
Both of which depend on the plpythonu functions running with plpython3u, in other words that they are Python3 compatible. Have you verified that?
> > *Solution 1:* > > After downloading and extracting the dump, convert the pgdump file to > sql file which is editable. > > * pg_restore -f out_dump.sql dump.pgdump* > > Replace all the plpythonu references with plputhon3u. > > Restore using the sql file. > > * sudo -H -u postgres psql -p 5433 -d db_name < out_dump.sql*
I would suggest working on the schema portion separate from the data:
pg_restore -s -f out_dump_schema.sql dump.pgdump*
Do your search and replace, restore to database and then:
pg_restore -a ... dump.pgdump*
Where -a is data only.
In fact if you have control of the pg_dump break it into two parts:
pg_dump -s ... --schema
pg_dump -a ... --data only
> > > *Solution 2:* > > After downloading and extracting the dump, get the list of items in dump > (Schemas, tables, table data, Index, functions, etc). > > * pg_restore -l dump.pgdump > dump.txt* > > Delete all the function references which have plpython3u.
I'm guessing you meant plpythonu above.
> *Question:* > > Our database size is 500GB, > > Do we see any performance impact using solution 1. Since solution 1 is > using sql file load and solution 2 is using pg_restore directly. > > Kindly recommend what to choose, solution 1 or solution 2 or any other > workaround to restore.
Personally I would go with solution 1 with the modifications I suggested.
