I have 20 pieces of dumpfile from a full export of a PDB. The total size of the dumpfiles is 5GB.
When imported to the new PDB, the size of the database is 150GB. I used the full import parameter below:
impdp system/password@test_pdb directory=DATA_PUMP_DIR dumpfiles=expdp_full_%U.dmp parallel=20 ...
It took 5 hours and 32 minutes to complete the import. What I noticed is loading the master table takes 1 hour.
I am doing import in a very powerful machine (Exadata). My database parameters are:
cpu_count = 256
sga_target = 250 G
pga_aggerate_target = 120 G
DB version is 19.3 (2-Node RAC with Dataguard)
(I am using ASSM)
Below is the portion of import logs I am pertaining to. Also, I noticed it takes time to load statistics and statistics marker, so I plan to exclude that and just gather fresh statistics. And finally, a lot of table partitions being imported have 0 rows. I think there are about 8000 plus partitions with 0 rows. They are being imported pretty fast though.
15-DEC-20 01:09:43.378: W-1 Startup took 1 seconds
15-DEC-20 01:10:22.316: W-1 Master table "SYSTEM"."IMPDP_FULL_J01" successfully loaded/unloaded
15-DEC-20 02:10:08.736: Starting "SYSTEM"."IMPDP_FULL_J01":
15-DEC-20 02:10:11.934: W-1 Processing object type DATABASE_EXPORT/PRE_SYSTEM_IMPCALLOUT/MARKER
15-DEC-20 02:10:32.251: W-1 Completed 1 MARKER objects in 3 seconds
15-DEC-20 02:10:32.251: W-1 Completed by worker 1 1 MARKER objects in 3 seconds
..
..
15-DEC-20 12:27:45.247: Job "SYSTEM"."IMPDP_FULL_J01" completed with 25 error(s) at Fri Dec 15 12:27:45 2020 elapsed 0 05:32:23