Added logging for the invalidation of diskCopies due to overwrites in another stager. Also fixed missing gcType in several places

Changed 'path' of files within data pools so that they do not have the extra directory that is the fileid%100

Rewamped deletediskcopy command. Mainly dropped the ability to restage and adapated to new DataPools

DataPools are pools of data, that is DiskCopies are directly placed in the pool, without any notion of FileSystem.
DiskServers can be associated to such a pool and are thus able to server all the DiskCopies present in the pool (several diskservers can serve the same DiskCopy).
The Schema has been modified and all the SQL code has been adapted to the new concept.
The change should be backward compatible, in the sense that regular DiskPools are still working the same way as they did before.
Note that DiskPools and DataPools can be mixed within a given service class but that there is no optimization of such a setup. This setup is more foreseen for transitionnal periods where optimization is a secondary issue.

Dropped unused entry point getBestDiskCopyToRead. This was used by xroot in long gone non scheduled mode

Fixed handling of nsOpenTime in new stager code. This was preventing recalls and d2d copies to work fine

Related to commit 2dc97f5c: optimized selection of best candidates for the recall, only taking into account filesystem ranking. On the way, dropped the function-based index computing the rate as its main use for the migration case has been replaced by a dual-level sorting, and the filesystem rating is executed directly

Fixed problem of the last answer to client that is still special as it triggers the closing of the connection

First step of a refactoring of the stager code, targetting the user file requests (get/put and prepareToGet/Put). The refactoring essentially consists in simplifying the complex structure of helpers and handlers and concentrate the code into the SvcThread classes. On the way, the update and prepareToUpdate requests have been dropped.

affecting draining operations.

Consolidation of the Disk2DiskCopyJob table: replacedDcId is gone
and srcDcId is used in all cases. To know whether the source diskCopy
is to be dropped, the dropSource (boolean) column has been added.
Moreover, FK constraints and indexes have been added where appropriate.

Extended deletediskcopy to deal with entire mount points.
This included retrieving the Cns_seg_metadata.status data
in bulk at the beginning for efficiency reasons.

One more improvement after a campaign of stress test: avoid to
signal() when someone else already did it. This helps reducing
contention on SYS.DBMS_ALERT_INFO - and ought to have been provided
by the Oracle DBMS_ALERT API...

Avoid to loop on DBMS_ALERT.SIGNAL on bulk operations. This involves
both the request handler on bulk PrepareToGet requests and the
Disk2diskCopyJob bulk handling.

CastorFile.nsOpenTime. Actually, the goal for next non-transparent
intervention should be to rename selectCastorFile() into
openCastorFile(), so that its scope is well defined, and the other
stager operations not needing an open() should just execute their
dedicated procedures. This should be in the scope of completing the
refactoring of the stager.

This was actually triggering deadlocks in production (CMS) with users
using non normalized paths. The deadlock was due to the autonomous
transaction in dropReusedLastKnownFileName() attempting to take a lock
on a CastorFile already locked by the main transaction, which is
forbidden.

  - avoid looping too much in the tapegateway when some file is locked for long
  - improved handling of log messages when the DLFLogs table is too full :
    + do not take all of it in one go
    + use an index as the table may be scattered

takes a lock, thus in case we need to signal we have to commit before hand.

Fixed check for ongoing prepareToPut when a new one comes. This could lead to concurrent prepareToPuts

Fixed race condition allowing 2 prepareToPut to start concurrently, due to a wrong check of the SubRequest status in the corresponding check

Fixed the timing of setting the nsOpenTime of the CastorFile : it needs to be done after all checks so that cancelation of the opening does not result in changing the nsOpenTime