Personal tools
You are here: Home → Documentation

Home Download Button
 

Release News

Release Candidate 1.0

May 6, 2008

Improvments:

  • Added options to vault descriptor file: Vault descriptor template is more complete (all options included but commented)
  • Added support for multi-threaded Slicestor. A bug in Mina prevented this from occurring before.
  • Added support for configurable thread pools. Both accesser and slicestor now have configurable thread pools. See properties.xml in each respective application.

Fixed bug:

  • Performance in cases of high network latency and multiple vaults per accesser is improved.
  • Fixed bug: Accesser daemon does not fail if port 3260 already in use
  • Fixed bug: (many related fixes) OutOfMemoryError and Out of swap space
  • Fixed bug slicestor-repair: Repairing a Block single file vault log issue
  • Fixed bug SCSI INQUIRY response indicates obsolete task management model state
  • Fixed bug Slicestor-Repair fails with server config files with DAEMON_ID
  • Fixed bug IDA and other codecs' encoders & decoders may be instantiated more often than necessary
  • Fixed bug Write performance degrades severely in the presence of network latency
  • Fixed bug ReadCapacity10 returns invalid response
  • Fixed bug Authentication to slicestor occasionally fails in slicestor test mode
  • Fixed bug Some implemented mode pages were not detected
  • Fixed bug Upon creating 25 vaults and targets and concurrent r/w causes "OutOfMemoryError: unable to create new native thread"
  • Fixed bug Slicestor starts using swap memory when 4 concurrent vaults are writing to the dsnet (1 initiator & 1 accesser)
  • Fixed bug Performance tests for concurrent vault upload has a very high CPU utilization.
  • Fixed bug SliceStores are not deleted after a failed vault creation.
  • Fixed bug iSCSI target should ensure that incoming task ITT is not already outstanding
  • Fixed bug All CLIs report incorrect error when connection can't be establihed
  • Fixed bug Rebuilder message unnecessary
  • Fixed bug Error Message - Grid transaction error (possibly recoverable)
  • Fixed bug getTransactionId() called after commit or rollback causes NullPointerException
  • Fixed bug: Corrected NullPointerException caused by invalid operational text key comparison during login phase (blockdevice-org rev 741)
  • Fixed bug: Fixed illegal reference to static fields in enum constructors (iscsi-org rev 22)
  • Fixed bug: Fixed bug where valid flag would always be set for sense data even when the information field is not valid (for fixed format sense data). (iscsi-org rev 31)
  • Fixed bug: This build was cut as an emergency response to DE536 which turned out to be more severe than previously thought.
  • Fixed bug: Removed libdbus from all builds due to license conflicts.
  • Fixed bug: Fix the statistics Fixed bug: they had become out of date due to core code changes
  • Fixed bug: Fix to not call Thread.interrupt() for grid tasks.
  • Fixed bug: Altering target unloading behavior, now canceling any writes in progress and calling ABORT_TASK_SET on scsi target to purge task state at time of client disconnect
  • Fixed bug: Implemented fix for problem where errors during SCSI tasks would write a response to the grid even if the task had already been aborted. Also changed AbstractTask to make Thread.interrupt called only for those task implementations that support it.
  • Fixed bug: Small work around for the buggy open iscsi initiator which unexpectedly drops connections on discovery
  • Fixed bug: Fixed TMTask handleRequest() to build proper Nexuses before calling execute() to the SCSI layer. Previously the LUN was not being specified (the Nexus constructors was being used incorrectly).
  • Fixed bug: Changed abort implementation in AbstractTask and DefaultTaskSet to immediately remove aborted tasks from the queue. Also fixed bug where responses would be written to the initiator for sense exception tasks even if they had already been aborted.
  • Fixed bug: All task management function requests made on the iscsi initiator are sent to logger as warnings.
  • Fixed bug: Changed ReadGridTask and WriteGridTask to retry operations on IO errors. Tasks will now retry until aborted by the initiator. This should cut back on spurious disconnects on faulty networks.
  • Fixed bug: Modified WriteGridTask to retry for a random time period between 1 and 5 seconds instead of the previously hardcoded 5 seconds in case of write failures. (blockdevice-org rev 862)
  • Fixed bug: Improved logging for read/write tasks. Read/write tasks now output the LBA and the number of blocks for each respective task at debug level.
  • Fixed bug: When an ABORT_TASK occurs, information about the aborted task is logged.