Insert many millions of lines coming from text files

[mike9]

Is this PK an number which will/should always increase?

If YES you might do the following:

Create an external table (source)
Find out the min and max value from the pk of the source
Using these 2 values you then could do an:

insert /*+ append */ into target
as select s.*
from source s,
(select pk_id from target where pk_id between MIN_SOURCE_PK_ID and MAX_SOURCE_PK_ID) t_pk
where s.pk_id = t_pk.pk_id(+)
and t_pk.pk_id is null

insert /*+ append */ into duplicates
as select s.*
from source s,
(select pk_id from target where pk_id between MIN_SOURCE_PK_ID and MAX_SOURCE_PK_ID) t_pk
where s.pk_id = t_pk.pk_id