PresSTORE Release Notes

• Fixed event processing for Sync runs.
• Improved recycling in single drive.
• Corrected handling of non-ntfs file systems on Windows.
• Corrected Windows installation – the not installing microsoft redistributable libraries.
• Improved Windows handling of lookup indexes.
• Improved job finalization on Windows, rename of index directory.
• Corrected saving the empty mount point.
• Corrected deleting deep direcotries upon full sync on Windows that presented itself as a connection abort.
• Still needs correction - remote installing Windows. The installation runs through but does not automatically restart.

• Fixed sending test mail from within user preference.
• Added new SUA users in the supported licenses database.
• Fixed protection of resource groups in the configuration database that prevented non-root users to submit restore and/or archive jobs.
• Added QuickStart help item which attempts to simplify first-time experience when evaluating the product.
• Automatically recycled volumes are not set to Suspect.
• Volumes blocked by the current cycle are released after a new full backup.
• Fixed daylight time saving for scheduled backup jobs.
• Fixed licensing of virtual disk devices within a jukebox so they dont steal license units from real media drives.
• Filter volumes marked for deletion from the volume list.
• Fixed taking inventory after recycling volumes.

• Fixed memory leak and potential memory trash in Mac OS X acl_to_text() library call as it is broken since 10.4 up to the 10.5 OS X version.
• Fixed race condition in backup indexer when two or more job executors may deadlock when attempting to release the indexer thread.
• Disabled error-counters inquiry for HP Ultrium-4 devices as some of them have broken firmware.
• Fixed broken Intel/Solaris distribution which failed to load on some Solaris10 buidls.
• Mail sender/server setting done in the configuration file are now automatically joined in the configuration of the localhost machine, so they are visible in the GUI.
• Added support for relocatable Desktop databases for Helios EtherShare fileserver.

• Fixed start-at-boot on Mac OS X broken by the previous patch release while moving startup machinery to launched.
• Fixed memory leak when saving or restoring files with total path length larger than 200 bytes.
• Improved restore selection to early skip files located on recycled volumes.
• Fixed media reservation problem that caused parallel saves to fail under some circumstances.
• Back-ported disk driver from the V3 including new faster compression algorithm and automatic volume-repair option.

• Fixed hangs in Sync control channel, happening with some high-speed and high-latency WAN links.
• Reduced timeout while establishing reverse client connection from 86400 to 120 secs in order to quickly recover from temporary client problems.
• Switched to use launchd on Mac OS X 10.5.
• Fixed preview generation using QuickTime on Max OS X 10.5
• Fixed inventory in jukebox without a barcode reader when volumes have not been previously labeled
• Prevents TCP socket leaking on reverse client connections during sync.

• Fixed backup task scheduling for plans saving more than one client, where one client backup runs longer than one day, whereas other client backup is done fairly quickly. In such cases the quicker backup may be re-scheduled using wrong startup time.
• Fixed remote agent timeouts when generating large Sync cycles.
• Synthetic backups now use unnamed pipes instead of TCP socket pairs to join restore and backup jobs. This speeds up the process and consumes less system resources.
• Added Mac OS X (Intel) support for Xinet desktops.
• Fixed download of files outside fileserver volumes on Mac OSX.
• Extended "Job xmlticket" CLI command to allow one optional argument specifying the path of the file where XML job ticket data should be stored.
• Prevent server restarts if there are very long running script executions underway.
• Fixed problems related to scanning file systems with exotic characters on Windows.
• Fixed problems related to index corruption when running an incremental backup on localhost on Windows.
• Do not act on hard read/write errors as reported from tape drives. Instead just log the event in the system log.
• Fixed SCSI device recognition on SuSE 10.9 Linux as the format of the /proc/scsi/scsi report changed.
• Avoid using "ConsoleMessage" alias on Mac OSX 10.5 (Lepopard) in startup scripts, as this isn´t defined any more.

• Added additional checking for session-ID in HTTP sessions between the clients and server, suppresing any potential session-ID reuse.
• Added one-time retry on communication errors during synchronize.
• Fixed recognizing users belonging to admin group on Mac by additionaly iterating over all known groups, instead of just fetching the group in question by getgrnam() library call. It seems to deliver incomplete list of group members in some rare cases.
• Fixed locating data in index on Windows 64bit installations.
• Fixed synchronizing and resynchronizing paths with special characters on Windows.
• Fixed closing volumes from different clients of the same job when "Use new volumes" is chosen.
• Fixed proxy command server when running commands for usernames longer than (or equal to) 6 characters. This resulted in some files not being deleted although archive jobs were configured for file deletion.
• Fixed handling of double-fork Xsan Mac files when the corresponding resource-fork could not be opened for reading or otherwise recognized. This resulted in files not being saved nor during incremental backups.
• For backup jobs running on 2 or more devices, the total job status is set to "exception" when at least one of the job executors ran in error. Only if all job executors finished without errors, the job status is set to "completed".
• Added "cli::index::backup::synchronize" CLI command to sync the backup index of the remote client with the actual state of the filesystem in cases where files deleted from the filesystem were not correctly marked in the index.
• Download of support data now includes all compressed copies of system.log file on Mac OS X.
• Fixed volume import of volumes created using synthetic backup.

• Fixed display of dataset sizes larger of 4TB during backup, synchronize and/or restore operations.
• Fixed duplication of archive, backup and sync plans not to duplicate the handle of scheduled job and thus prohibiting schedule and startup of the newly duplicated plan.
• Fixed compatibility with Internet Explorer V7 on Windows.
• Simplified handling of desktop database update for Helios servers. In some cases during the sync when files are changed and copied to the target directory, we unnecessarily adjusted the ID of the parent directory which sometimes lead to dup IDs.
• Filter out system and library cache and virtual memory files on Darwin so they do not get synced, archived or backed up.
• Fixed preview generation on remote clients when path to preview file contained "+" character.
• Virtual jukebox wizzard adds two media devices per unit instead of just one, thus allowing synthetic backups to be run on one jukebox unit.
• File download always generates Mac OS X compatible zip files.

• Fixed open-file leak when archiving Xinet volumes containing bogus (zero-length) resource files.
• Enforce cloned volumes are always written in pairs even when the pool has been re-configured to disable cloning.
• Added automatic repair of inproperly closed disk-volumes.
• Changed handling of connections to Helios desksrv. Instead of testing the connection with a dummy GetDeskCharset() call which proved to be problematic and was causing the desksrv to freak out sometimes, we now deliberately run into error when running against a stale connection. On certain set of errors we re-try the connection to desksrv and re-try the failed operation once more.
• Fixed reserving the oldest written volume for writing first, in case drives are explicitely assigned to the pool.
• Added new DLT device types to choose from when installing a changer.
• Fixed job stophours processing when the very first backup of a plan was cancelled.
• Fixed job auto-cancellation timing when no job monitor was running during job execution.
• Added CLI tests for measuring network speeds.

• Fixed ACL handling for directories during sync and restore.
• Changed handling of connections to Helios desksrv. Instead of tearing down the connection each idle seconds, we cache the connection indefinitely. On re-use, the connection is tested and if stale, newly established.
• Corrected job re-scheduling in between switch to and from daylight saving time.
• Corrected suppressing recycling of volumes when a backup job backs up more than one host.
• Corrected Windows deadlock during thread synchronization for multiple jobs.
• Fixed jukebox handling for some Spectralogic units that failed to report a correct element-move-matrix.
• Fixed starting the software under non-privleged user.
• Supressed errors about new/existing symbolic link existence on the target of a sync plan when running on Mac OSX.

• Resolved conflict with different file-type during the sync by making a new version of the target file.
• Sync with delete will overwrite any exiting target files whose modification dates differ from the source.
• Resolved deadlock on archive databases when one user was in the middle of search query preparation while some other was about to mount the database.
• Fixed deadlock during synthetic backup when more than one client is simultaneously backed up by the same plan.
• Removed localhost check done by probing the IP address. The localhost test is now done solely on the base of the match on "localhost" or result of the "hostname" command.

• Fixed broken client-to-server synchronize direction introduced with 2.3.6 release.
• Avoid saving empty index chunk to media device if the save job is prematurely cancelled or otherwise broken.

• Added selectable type of full backup for the monthly schedule. It now allows selection of the synthetic type as well.
• Fixed download of support data from Windows servers.
• Fixed link to server logs for the access from the GUI on Windows platform.
• Added configurable timeout for remote-tree mounter.
• Persist loading application resources from zip archives if failed to serve them from the pre-built cache.
• Helios driver allows reuse of file/folder ID during the sync or restore if the target volume is set to read-only.
• In CLI, allow setting of volume label/location/state/mode parameters, in addition to getting the value.
• Added condition log that tracks all state changes for the lifetime of a volume introduced by the application code.
• Added GUI warning when about to delete a volume.
• Added automatic (implicit) clone-volume re-synchronization.

• Added utility to test and rectify potential wrong setup of jukeboxes containing 2+ drives. Currently only comand-line (CLI) access is opened.
• Fixed indexer bug that sometimes caused overlay of directories and shadowing (make invisible) of regular file entries after incremental backups were performed.
• Synthetic backup does not conclude written volumes if the same pool is used for full, increment and synthetic. We now split volumes needed for reading from those required for writing at pool mount time.

• Added environment variables passed to pre/post script.
• Corrected recycling of imported volumes, which could not be relabled without deleting them from the volume list first.
• Removed setting of clone volume callbacks when starting a job, which caused the chunk block to be empty and consequently making the cloned volumes not importable.
• Changed the type of strictStreams property from boolean to integer. No consequence and not visible to the user. Needed so as to limit the number of clients per stream.
• Added limiting the number of clients per stream. This is still not opened thru the GUI.
• Fixed handling of IO-Registry path on Macintosh devices as saved in the configuration database if the path contained blanks.
• Changed SCSI message handling on Mac OSX Intel platform
  Under some circumstances we could trigger a bug in the SCSI Task kernel driver and cause the machine to panic.
• Avoid writing index on tape devices which were being SCSI-bus resetted in the middle of write operation.
• Fixed loading of some AIT drives by waiting for extra long time for media to be fully loaded.
• Improved volume resevation for synthetic backup, removed conflicts between volumes to read and to write.
• Fixed broken submit/start of backup plan over the CLI.
• Forbid creating resources with special reserved names.
• Forbid creating reserved user preference records at login.
• Added Adobe Photoshop files to list of supported image formats for Quicktime image conversions.
• Fixed autoinventory when jukebox is being used by some other job.
• Improved job monitor messages when volumes or their clones are requested.
• Fixed creation of the first file in a directory on windows when it´s parent directory path is missing and needs to be created.
• Fixed restoring from a mixed set of volumes and their clones.
• Improved autoinventory to check only the devices assigned to the given pool.
• Added additional cleanup and parking the medium in case the follower failed to load.
• Added setting the medium to suspect in case we expected to find a label, but found the medium to be empty instead.
• Changed to always use the available primary media, before using the clones.
• Corrected reporting of volumes used for restore when the job monitor is not opened.

• Fixed jukebox and drive installation wizard on Darwin.
• Corrected reporting of used volumes in job log.
• Fixed deadlock when two or more jobs are terminating at the same
  (or roughly the same time).
• Include volume and its clone in the list of volumes when per-volume browsing
  of the archive database.
• Fixed populating of restore basket from backup index.
• Fixed cloned volume reservation when volumes are in seperate jukeboxes.

• Removed spurious mail about device not available when changing
  media during a save job on Sun.
• Removed breaking a job when the hard read/write flag was raised but
  otherwise no error occured.
• Fixed GUI browsing of the backup index which failed to properly
  address elements in the index, resulting in empty root-level
  directory contents.
• Fixed automatic recognition of special directories (fileserver
  volumes) when the above (parent) directory is used as the source (and/
  or target) in a sync plan.

• Changed parsing of script and arguments as entered in pre/postscript fields in sync, archive and backup plans. The new parsing scheme requires tokens to be divided by spaces and if the tokens themselves contain spaces, they have to be enclosed in curly braces.
• Made fixed archive directories sticky so they don´t get deleted after the archive, if the file deletion was configured.
• Fixed re-schedule of Sync plan when just one runtime has been entered and removed from the scheduler.
• Fixed delete pass for Sync when cycles are configured. Also, the delete pass now purges the target directory from versions.
• On Windows, do not abort file walk if we cannot open some directory. Instead, log the event in the server log and continue.
• Fixeddeadlock occuring when multiple threads start to log and one of thethreads is in the middle of updating the configuration database.
• Changed semantics of hard read/write tape errors. In case one of those is encountered, the volume is set to suspects tate, email to administrator is generated but the operation underway is not aborted. The reason for this is that some broken tape-firmware erroneously report hard errors after some absolutely legal operations.

Version 2.3 released. Changes since last release

• Added support for driving SCSI devices like tape drives and jukeboxes on the Windows platform.
• Added QuickTime still-image, film and audio previews for the archive module. This is currently only supported on the Macintosh platform.
• Added support for Netatalk, an open-source file andprint server for Macintosh computers running on Unix servers. This is currently only supported on Linux platform.
• Replaced socket command daemon (scd) with the newproxy slave(s) processes operating over pipes. This applies to Unix-like operating systems only.
• Added new extended job log containing all relevantlog messages from the server log file pertaining to that specific jobonly. The generation of the extended log is controlled with"jobProtocol" global configuration option of the job scheduler module. The default value is "true" which turns-on the extended log generation.
• Added "maxConcurrentJobs" global configurationoption for the job scheduler. The default is set to 16. To change thevalue, for example to 32 concurrent jobs. add following two lines tothe configuration file conf/lexxsrv. :
  ns_section ns/server/$server/module/jobsched
  ns_param maxConcurrentJobs 32
• Added "testInodeChangeTime" global configuration option for the indexer so wecan ignore or honour file ctime modifications when selecting files forprocessing. The default value for the option is "true" which includesinode change-time checking. To ignore ctime checking, add followinglines to the configuration file conf/lexxsrv. :
  ns_section ns/server/$server/module/bixindex
  ns_param testInodeChangeTime false
• Addedper-plan email addresses. If configured, job reports are sent to thegiven email addres(es) instead to the job owner (i.e. the user whostarted the job).
• Manual start of backup plan does not automaticallytrigger execution of follower plans. Only the selected plan is executedwith the selected level.
• Added per-user preference to stay logged-in in theWeb-GUI even when the browser has been terminated without explicitlylogging out.
• Changed click behaviour in browser to allow double-clicks and right-button-select on list elements.
• Fixedincorect handling of bucket pool caused by the new signalizationcapabilities between the peers during the data trasnfer. This couldeventually drain the entire bucket pool and bring the data-transfer tohalt. Fixed compression handling for Sony AIT4 tape devices with therevision 1.0.3 or earlier. The tape device just went astray when weattempted to force-set the compression flag. Customers are encouragedto update to the latest firmware release (1.0.4 as of 29.08.06)

Version 2.2.4 released. Changes since last release

• Adjusted AppleDouble parser to cope with zero-length entries as created in meta-files on Apple´s Xsan.
• Fixedbug in parsing synthetic plan value which was introduced by fixes in2.2.3 release and rendered synthetic plan usage impossible.
• Fixed protection checking on non-root restoreswhere we wrongly checked the parent´s parent directory instead of justparent directory of the file being restored.
• Fixed display of values for newly configuredarchive plan where default values have been reported, instead of thenewly entered ones.
• Increased default timeout value for caches from 3 to 60 secs.
• Fixed GUI bug which prevented volume import from a standalone device.
• All resource deletions from GUI are now logged into log file.
• Automaticallycreate .HSancillary files on Xinet volumes because Xinet´s own librarydoes not do that which results in potential loss of Apple´s FinderInfostructure.
• Allow for passing additional arguments to pre/post scripts.
• Fixed case where syncplan was using the delete flag wrongly.
• Improved backup index browsing to properly select the right index chunk when more than one plan is saving the same tree.
• Added option of overriding built-in email delivery by means of a user-supplied script. For more information please refer to: http://www.archiware.com/wiki/wiki/96

Version 2.2.3 released. Changes since last release

• Fixed root of the archive job when more than one entry is selected for archival.
• Fixed protection issue when non-root users were restoring files directly on top of the root filesystem.
• Fixed synchronize of files larger than 4GB.
• Re-arrangedlayout of AppleDouble files generated on non-HFS filesystems in orderto overcome finder crashes introduced by over-optimistic Apple´soptimizations which broke the AppleDouble standard. Also, skipsgenerating the AppleDouble meta-file if the file contains no metadatanor resource-fork.
• Fixed rare cache-flush errors on configuration database.
• Automatically stop scheduled jobs for all backup tasks belonging to a certain backup pan when the backup plan gets deleted.

Version 2.2.2 released. Changes since last release

• Fixed drive reservation when large number of jobs are using same pool resources
• Fixed cross-platform Sync (Helios/Solaris -> MacOSX/HFS+) where synced files got wrong modification date
• Temporarilyremoved ACL/metadata handling for cross-filesystem (HFS<->Xsan,HFS<->UFS) operations because files containing those parts causedFinder to crash and restart. The bug in Finder has been reported toApple for further examination.

Version 2.2.1 released. Changes since last release

• Fixed incremental backup for windows clients.
• Fixed broken layout of apple-double files on non-hfs filesystems which contained metadata values.
• Speedup of sync jobs between Mac OSX computers which operated on large number of directories.
• Fixed searches in backup indexes which had to traverse more than one index cycle.
• Fixed occasional "peer closed connection" errors from clients.
• Changed handling of timestamps of Xinet resource forks. They are now set to the modification time of the data fork explicitly.
• Fixed broken automatic jukebox inventory.

Version 2.2 released. Changes since last release

• Added support for MacOSX Intel platform
• Added more options to user preferences to control mail delivery
• Job monitor shows up to 10 last jobs in list of jobs of today
• Fixed update of archive database(s) after volume migration
• Socketdriver now listens at IP address 0.0.0.0 which includes all activeinterfaces and bypasses reverse DNS IP-to-hostname queries, whichresulted in server starting problems in some wrongly configured IPenvironments
• Fixed parsing of file paths in sync plans which contained blanks
• Added note about follower volume during volume migration
• Fixed CLI jukebox inventory command to be compatible to 1.5 version
• Backupindex view does not prune deleted directories. Only deleted files arepruned from the index view when snapshot is selected
• Backup index does not limit the view to the lastincremental run, rather it shows all files and folders unfiltered. Onemust use index browse options to setup specific visit-time to get thesnapshot view
• Fixed handling of Windows reserved path characters when working against directories under WSM control

Version 2.1.0 released. Changes since last release

• Added support for Solaris x86 platform
• Added configurable timeouts for remote command sessions
• Added new tape-media types
• Added flag to use new media for volume migration
• Extended command-timeout to remote clients if media mounting takes extra long time due to lack of system resources
• We do not attempt to open resource forks for non-regular files on HFS filesystems any more
• Fixed cancellation of sync jobs from server to client
• Fixed labeling of media in standalone drives
• Fixed transfer byte-count calculations which was causing syncs of large directories (large number of files) to fail
• Fixed alternate (user-given) index root for archive jobs
• Fixed integer overflow in summary line when reporting total size of data saved accross 2 or more drives
• Fixed rare server cores happening when clicking on a new job log entry in job monitor
• Fixedincremental backups saving the same modified files for each invocation,thereby creating unnecessary index entries and wasting tape storage
• Fixed parsing of bad start times in backup and archive plans
• Fixed handling of backup index root directories containing special characters in pathname
• Improved progress reporting during volume change on restore

Version 1.5.14 released. Changes since last release

• Fixed job executor blocking on very large number of parallel executing jobs.
• Fixed searching in archive indexes when very first reported hit is located on non-existent volume.
  In such cases we got in endless loop effectively blocking the access to the database being queried.
• Added protective locks arround Xinet desktop library which seems to be broken in MT-respect.
• Added support for Sony PDD devices on Mac OSX and Linux.
• Added adjustable session timeout settings for CLI sessions. The default is to expire the session
  after some inactivity time is passed. The inactivity time depends per default on the time of the day
  and varies between 2 hours and 1/2 hour.
  To override default setting, change to PresStore installation directory(default = /usr/local/aw), edit conf/lexxsrv. file, where is the TCP port for the PresStore server (default = 8000)
  and add these lines at the end of the file:
  ns_section ns/server/$server/module/clisrv
  ns_param sessionTimeout
  is the number of seconds for the timer.
  Possible values are:
  0 - default setup
  >0 - custom timer value
  -1 - timer is disabled (session never expires)
• Fixed pre/post script execution on client which seemed to be broken since some earlier patch-release.
• Loading media from mail-slot implicitly starts first barcode then mount inventory of the newly loaded
  slot (optionally).
• Sync plans configured to check compatibility (default) between the source and the target directory
  refuse to operate if during the runtime the target directory is missing.
• Fixed proper volume mounting of (already) burned DVD-/+R(W) media when restoring data from
  multiple mixed-media volumes (a combination of DVD-RAM´s and DVD-/+R(W) media).
• On Mac OSX, added 10 minute alarm signal in the watchdog proces to periodically test the server
  process existence and restart if the server died w/o properly signalling the watchdog. This behaviour
  was observed since 10.4 (Tiger) release and resulted in server process not properly restarting when scheduled.
• Fixed metadata handling in the file walker thread which caused occasional process hangs and/or cores.
• Fixed configuration database recovery when the restored database schema did not match the
  schema used by the currently deployed software version.
• Circumvented problems at media load for some revisions of IBM LTO3 tape-drives. Instead of
  immediately reacting on insane error codes delivered by the drive, we repeat the operation up to
  4 times with 2 seconds delay in between and hope to get the drive finally load the media.
• Mount inventory on optical jukeboxes with empty caddies failed prematurely. Instead of breaking
  the job at the first empty caddy, we count this case as valid, report the event and continue with the operation.
• Added support for Linux Debian 3.0+ release.
• Fixed sorting of archive selection when feeded with paths t