1. 03 Sep, 2014 1 commit
  2. 01 Aug, 2014 1 commit
    • Giuseppe Lo Presti's avatar
      Extended diskmanagerd to take over the role of the stagerjob. · d69c9c41
      Giuseppe Lo Presti authored
      This includes:
      - a new thread to reply to clients using the CASTOR2 protocol (cf.
        stager's RequestReplier).
      - a new thread listening to clients coming back to start the transfer.
        This thread polls all outstanding client connections and just accept
        them, so to pass the file descriptor as stdin/out/err to the mover
        (inetd mode).
      - a new thread listening to movers for their file close
        operations.
        Details on the implemented protocol in moverhandler.py.
      - Extra exposed_ methods to handle the meta-data operations.
        Those replace the stager Job service, which is now obsolete.
      d69c9c41
  3. 25 Jul, 2014 1 commit
  4. 01 Jul, 2014 1 commit
  5. 05 Feb, 2014 1 commit
  6. 01 Aug, 2013 1 commit
  7. 12 Jun, 2013 1 commit
    • Sebastien Ponce's avatar
      Refactoring of the disk to disk copy code. · 9bb2b135
      Sebastien Ponce authored
      This includes :
        - merge of d2dtransfer into diskmanagerd
        - drop of the WAITDISKTODISKCOPY state in DiskCopy, and of the diskcopies for ongoing replication in general. Now DiskCopies are only created at the end of the replication, when they actually exist on disk
        - drop of the StageReplicaRequest table and concept
        - introduction of the DiskToDiskCopyJob table, holding the list of ongoing disk to disk copies
        - a split of transferToSchedule into userTransferToSchedule and D2dTransferToSchedule
        - an according split in the dispatcher part of the transfer manager to call both methods. Note that with this commit, both methods fill the same FIFO queue of transfers and both take everything they can find without any throttling.
      
      At this stage, the draining facility is broken.
      9bb2b135
  8. 28 Mar, 2013 1 commit
  9. 04 Feb, 2013 1 commit
  10. 10 Dec, 2012 2 commits
  11. 27 Jul, 2012 3 commits
  12. 24 Feb, 2012 1 commit
  13. 07 Sep, 2011 1 commit
  14. 16 May, 2011 1 commit
  15. 13 May, 2011 1 commit
  16. 20 Apr, 2011 2 commits
  17. 15 Apr, 2011 2 commits
  18. 05 Apr, 2011 1 commit
  19. 01 Apr, 2011 1 commit
  20. 31 Mar, 2011 1 commit
  21. 29 Mar, 2011 1 commit
  22. 25 Mar, 2011 1 commit
    • Sebastien Ponce's avatar
      Many small improvements and fixes to the transfer manager and tools around it.... · 42bec4d2
      Sebastien Ponce authored
      Many small improvements and fixes to the transfer manager and tools around it. Most credits for finding the problems go to Dennis.
        - fixed header alignments in listtransfers
        - fixed case of keys in logs
        - set JobId to "unused for TM" in the stagerJob log
        - create the temporary files of the scheduling read only
        - show fileid in listtransfers
        - record the request id in all log messages where it makes sense
        - fixed rebuilding of diskserver manager queue when it restarts
        - handled properly log values contaning spaces
        - fixed too zealous retries to schedule d2d destination when source is not ready
        - improved drain mode, so that the transfer manager stops running when activity is over
        - fixed automatic reload for all parameters (some were cached and thus not reloaded properly)
        - reenabled active canceling of jobs on diskservers that did not start it first (was lost in some bad merge)
      
      42bec4d2
  23. 22 Mar, 2011 2 commits
    • Sebastien Ponce's avatar
      Many fixes after a big testing session · fa20a8df
      Sebastien Ponce authored
      fa20a8df
    • Sebastien Ponce's avatar
      Several improvements on the new scheduling component : · c4b41fad
      Sebastien Ponce authored
        - lot's of renaming. jobs are now transfers, client commands are named listtransfers, killtransfers and draintransfers, daemons are called diskmanagerd and transfermanagerd, packages are renamed accordingly, ...
        - a drain mode has been added to the transferdaemon to ease the retirement of machines running it
        - ports used are not colliding anymore with LSF so that we can let LSF run and only stop/start the jobmanagerd and the transfermanagerd when we switch from one mode to the other and back
        - the badmin command has dropped in favour of automatic, regular reload of all configuration files
        - client commands have been extended to allow restriction to a given diskpool and user
        - the listtransfer commands has been extended to display more parameters, depending on options (e.g. number of unique pending jobs in a pool, per protocols values)
        - reconnection to the ORACLE DB have been fixed
        - DLF logging has been fixed (both the wrapping of it through enums and the insertion of new facilities in the facility table)
      c4b41fad