Lowest SCN in an existing LogMiner dictionary

Question

What is the SQL query to know the lowest SCN in the existing Logminer dictionary that can be used as the FIRST_SCN of a NEW capture process?
Thanks.

Answer 1

Sound like, your sentence has been weighted as your wording poses problem, specially 'existing' and 'new' which restrict the context. For my answer, I will suppose that the proposed data to replicate does not exists at target site.
You would like to know how far back in time you could go to replicate source tables. In theory there is no limit. You could even instantiated the new target table to SCN 1 and load all the archives that ever existed on your DB, prepare this table with SCN=0, create empty structure on target and the new capture will mine all your archived logs.
Alas this is false. If you do this you will see on the target site alert log 'MULTI VERSION DATA DICTIONARY' nick named MTVD. For there is another player: the dictionary export into archived logs. Streams needs to validate any LCR against a 'fixed in time' version of data dictionary to guarantee past consistency (meantime, did you add a column?). This export is made by DBMS_CAPTURE_ADM.build export the current data dict into an archive logs ( done automatically when you set up a new capture with DBMS_STREAMS_ADM.ADD_(TABLE/SCHEMA)_RULES ).
So the real restriction to your request is the presence of a copy of past point in time data dictionary into an archived log. And this is also the answer to your request : the lowest SCN where v$archived_log.dictionary_begin='YES'
The export itself, which install a set of old data into the target, could be resolved by restoring a copy of the source DB at this time in past you want to start the replication and creating an export from this restored back-in-time DB.
But you can't do anything to enforce the presence of a data dictionary into a past archive log.
So once you have re-created a back in time export and spotted an old SCN in v$archived_log where dictionary_begin='YES' then you may load all, the archives starting of this archive, create your NEW capture, instantiate source and apply site to your export SCN (whose SCN is also comprised between FIRSTCHANGE# and LAST_CHANGE# of this archive with dictionary_begin='YES'.
But all this is true if it is a NEW capture, for if you want to reuse an existing capture, then it is a bit more complicate to inject old data of not-yet-ever-replicated-table you want to add.