'the SRM way': cf. commit ab537e9e.

The only valid way to ask for all locations of a given CastorFile is
by passing REQUESTQUERYTYPE_FILENAME_ALLSC as query parameter type.

This is a temporary revert of a commit that will not be released in the next
2.1.14-x release.  The original commit should be put back in a future release.

Revert "bug #103776: Recall mount creation is not selective enough, leading to empty mounts."

This reverts commit e7f278cd.

accounting is no longer supported.

Added a dry-run check in recall mounts creation, like for migration mounts to
protect against global discrepencies (including unavailbility of a whole disk
pool as once experienced in production during a network problem).

We just make sure the statistics for mount creation are made with
actually pending recall jobs.

Updated the example castor.conf file to reflcect the fact that the folloiwng
configuration parameter has been removed:

TAPE LOCATE_TIMEOUT 180000

Fixed in the v2_1_14Version branch.

Fixed in the v2_1_14Version branch.

Fixed in the v2_1_14Version branch.

Fixed in the 2_1_14Version branch.

Added a second order ordering to the "next tape for migration" logic of the
vmgr daemon in order to satisfy a request from Tim Folkes.  Please see the
bug report in Savannah for more information.

Fixed in the v2_1_14Version branch.

Again my code compiles on SLC5, let's lean on TeamCity and see if the same
code compiles on SLC6.

compression statistics works on SLC5 but not on SLC6.  I am therefore
reverting my changes.

Please not ethat this is temporary.

the prefix 'all:' to the CASTOR filename.
This is deliberately undocumented, as the query is reserved
for admin/test purposes (and it is protected with CUPV rights).

With this commit the logic replacing the deltpfile() function now only logs an
appropriate message and then allows the rtcp child process to end the current
tape session gracefully.

This commit removes the retry logic from the rtcpd daemon.  The retry logic has
to be removed before the replacement of the deltpfil() function can safely
not call exit(-1) and let the current rtcpd tape session end gracefully.

Unfortunately I had to lean on TeamCity and break the build in order to
discover my last commit did not include a vital modification to the
LogHelper class of the tapebridged daemon.

The current removal of the deltpfil() function is not satificatory.  The current
solution effectively replaces the deltpfil() function with a log message and a
call to exit(-1).  The call to exit(-1) prohibits the propagation of a TAPE
OVERFLOW error to the tapegatewayd daemon in the event that a tape server
writes past the physical end of a tape.  This side effect means that tapes are
mounted in a loop once they become physically FULL because the vmgrd daemon is
never notified that they are FULL.  The mount loop is actually finite because
depending on the length of the last file(s) be written to tape the TAPE OVERFLOW
is detected by different parts of the tape server code and the parts that would
not have called the deltpfil() function are able to propagate the TAPE OVERFLOW
error to the tapegewayd daemon which in turn is able to notify the vmgrd daemon
that the tape in question is now FULL.

The new idea that is yet to be implemented, is to simply log a message instead
of calling deltpfil() and then let the rtcpd tape session end gracefully.  This
graceful end of tape session must not run the risk of trying to write to tape
again in some form of retiry logic.  I am therefore currently trying to remove
all retry related code.  One such collection of related code is the now
obsolete abort protocol, where a client of rtcpd sends it an RTCP_ABORT_REQ
message.  This commit removes the defunct marshalling code for RTCP_ABORT_REQ
messages from the tapebridged daemon.

I have made the save_errno and save_serrno variables of the "horrible macro"
CHECK_PROC_ERR local.  These two variables should never have been grabbed by
the macro from the enclosing function tapeIOthread().

procedure recompileAll at the end of all upgrade scripts.
This meant declaring the procedure in all databases (yes, including
CUPV where there's no PL/SQL code...).

script that was never released.

Removed the deltpfil() function that deletes files from the end of a tape in
order to leave the tape terminated with a correct label structure.  The
deltpfil() function increases the risk of data loss in a system that uses
deferred tape flushes.  When appending data to the end of a tape the
deltpfil() function has been made redundant by the fact that the stageri
always instructs the tape system to position to the end of the last file
successfully written to tape as opposed to requesting a tape drive to position
to end of data.

submitted requests to disappear.
The logic to clean up old requests has been changed: failed requests
are kept respecting the timeout in CastorConfig, successful requests
may be dropped earlier (as in the past) to keep the SubRequest table from
growing too much.
In consequence, the repack reporting has been adapted: the 'migrated'
count is now computed by difference with the other counters, while the
'failed' count can be trusted because of the above.

later on.

options to DiskManager/*PendingTimeout to improve usability.

affecting draining operations.

Completed the fix to take into account disk-to-disk copies failed
at the start.
To have a full tracking, a src diskcopy ID field has been added
to Disk2DiskCopyJob, and populated at start time when available.

Fixed tight loop on errors when filesystems of the gcd are badly configured. This could go up to 10KHz !

However, this is to be evolved to a more robust schema, where
DISABLED hardware for too long is deemed not to have any transfer any
longer and won't stop the synchronization from being executed on the
rest of the cluster.