Now the export is performed for both tapepools holding the two tape copies,
and the process aborts if a single 2nd copy tape pool is not found
(cf. README.md for more details).

Related to d6f5b554, now the procedures perform some pre-checks
to ensure CTA can handle dual copies, and to optimize their migration

The optimization helps the subsequent EOS metadata import, as
tape pools holding 2nd copy files required a needless pass over
the already imported files metadata. This optimization assumes
that dual copy files are imported in order, i.e. tape pools
holding copy #1 are imported prior to tape pools holding copy #2.

Documentation updated accordingly.

- Make sure no parallel queries are executed, to privilege a single-transaction approach.
  Should this turn out to be too slow, the temp table must become persistent across commits.
- Moved the tool one level up for clarity as it is not interacting with CASTOR.

- Added tool to drop from the CTA catalogue all CASTOR tapes from a given tapepool.
  This tool makes use of a temporary table and an additional PL/SQL procedure
  in order to perform efficient bulk operations in the CTA catalogue.
- Made tapepool insertion idempotent.
- Updated documentation.

[migration] document and log constraint violation errors when importing tape files. We still have to automate the import of those "missing" ones.

[migration] added code to log all details of SQL errors such as constraint violations during the bulk injection into Archive_File and Tape_File

Following discussions, it appears that experiments will be able
to only change a fixed prefix, i.e. /castor/cern.ch -> /eos/cta<instance>,
and not the deeper path components.
Moreover, we need to call getPathForFileId() on missing paths,
therefore an extra synonym + grant is added here.

- The CTADirsHelper is now a view as opposed to a table, and
a new CTATopLevelDirsHelper table has been created. Grants
have been changed accordingly.
- The CTA schema also got a new synonym to access the real table
underlying the CTADirsHelper view in order to speed up one query.

This assumes that the ARCHIVE_FILE_CB1_NN constraint is lifted
before an import can be executed!

Giuseppe Lo Presti authored Jun 21, 2019 and Michael Davis committed Jun 24, 2019

CASTOR migration: alternate implementation to fill the directories intermediate table with no temp table

CASTOR migration: alternate implementation to fill the directories intermediate table with no temp table

The code now runs in parallel mode according to Oracle EM.