Patch-ID# 108261-11 Keywords: Netbackup 3.2 jumbo patch 326 260 363 413 466 521 veritas Synopsis: Netbackup 3.2 data corruption fix and jumbo patch Date: Mar/13/2002 Solaris Release: 2.5.1 2.6 7 8 SunOS Release: 5.5.1 5.6 5.7 5.8 Unbundled Product: Veritas NetBackup Unbundled Release: 3.2 Xref: Topic: Netbackup 3.2 jumbo patch Relevant Architectures: sparc NOTE: i386 BugId's fixed with this patch: 4244797 4267724 4268538 4268539 4268540 4268541 4268542 4268543 4268545 4268546 4268547 4268548 4268550 4282898 4286545 4286547 4287760 4292771 4303994 4314679 4323166 4340667 4341898 4345788 4370437 4394584 4397701 4499273 4507040 4510169 4637467 Changes incorporated in this version: 4637467 Patches accumulated and obsoleted by this patch: 108777-01 112066-01 Patches which conflict with this patch: Patches required with this patch: Obsoleted by: Files included with this patch: /openv/bin/vauth_test /openv/bin/vauth_testd /openv/bin/vauth_util /openv/bin/vopie_util /openv/bin/vopied /openv/lib/libvopie.so /openv/lib/server/Solaris/Solaris2.5/libsfr.so /openv/lib/server/Solaris/Solaris2.6/libsfr.so /openv/netbackup/bin/2GB_bpdbm /openv/netbackup/bin/add_slave /openv/netbackup/bin/admincmd/bpauthsync /openv/netbackup/bin/admincmd/bpbackupdb /openv/netbackup/bin/admincmd/bpccname /openv/netbackup/bin/admincmd/bpclassnew /openv/netbackup/bin/admincmd/bpclclients /openv/netbackup/bin/admincmd/bpcldelete /openv/netbackup/bin/admincmd/bpclexclude /openv/netbackup/bin/admincmd/bpclient /openv/netbackup/bin/admincmd/bpclinclude /openv/netbackup/bin/admincmd/bpclinfo /openv/netbackup/bin/admincmd/bpcllist /openv/netbackup/bin/admincmd/bpclsched /openv/netbackup/bin/admincmd/bpclschedrep /openv/netbackup/bin/admincmd/bpclschedwin /openv/netbackup/bin/admincmd/bpclvalid /openv/netbackup/bin/admincmd/bpconfig /openv/netbackup/bin/admincmd/bpcounts /openv/netbackup/bin/admincmd/bpcoverage /openv/netbackup/bin/admincmd/bpdbjobs /openv/netbackup/bin/admincmd/bpdrfiles /openv/netbackup/bin/admincmd/bpduplicate /openv/netbackup/bin/admincmd/bpenableLN /openv/netbackup/bin/admincmd/bpenableTD /openv/netbackup/bin/admincmd/bperror /openv/netbackup/bin/admincmd/bpexpdate /openv/netbackup/bin/admincmd/bpflist /openv/netbackup/bin/admincmd/bpfrag /openv/netbackup/bin/admincmd/bpgetconfig /openv/netbackup/bin/admincmd/bpimage /openv/netbackup/bin/admincmd/bpimagelist /openv/netbackup/bin/admincmd/bpimmedia /openv/netbackup/bin/admincmd/bpimport /openv/netbackup/bin/admincmd/bplabel /openv/netbackup/bin/admincmd/bplicense /openv/netbackup/bin/admincmd/bplocaladdrs /openv/netbackup/bin/admincmd/bpmedia /openv/netbackup/bin/admincmd/bpmedialist /openv/netbackup/bin/admincmd/bprdreq /openv/netbackup/bin/admincmd/bprecover /openv/netbackup/bin/admincmd/bpretlevel /openv/netbackup/bin/admincmd/bpschedreq /openv/netbackup/bin/admincmd/bpstuadd /openv/netbackup/bin/admincmd/bpstudel /openv/netbackup/bin/admincmd/bpstulist /openv/netbackup/bin/admincmd/bpsturep /openv/netbackup/bin/admincmd/bpsyncinfo /openv/netbackup/bin/admincmd/bpverify /openv/netbackup/bin/bp /openv/netbackup/bin/bpadm /openv/netbackup/bin/bpbrm /openv/netbackup/bin/bpdbm /openv/netbackup/bin/bpdm /openv/netbackup/bin/bpgp /openv/netbackup/bin/bpinst /openv/netbackup/bin/bprd /openv/netbackup/bin/bpsched /openv/netbackup/bin/bptm /openv/netbackup/bin/goodies/NCVU.README /openv/netbackup/bin/goodies/NCVU.tar /openv/netbackup/bin/goodies/bpend_notify.oracle_bli /openv/netbackup/bin/goodies/new_clients /openv/netbackup/bin/goodies/post_checkpoint_notify.oracle_bli /openv/netbackup/bin/goodies/xbpadm /openv/netbackup/bin/goodies/xbpmon /openv/netbackup/bin/goodies/xnb /openv/netbackup/bin/update_javaclients /openv/netbackup/client/Solaris/Solaris2.5/bp /openv/netbackup/client/Solaris/Solaris2.5/bpbackup /openv/netbackup/client/Solaris/Solaris2.5/bpbkar /openv/netbackup/client/Solaris/Solaris2.5/bpcd /openv/netbackup/client/Solaris/Solaris2.5/bpclimagelist /openv/netbackup/client/Solaris/Solaris2.5/bpclntcmd /openv/netbackup/client/Solaris/Solaris2.5/bpdynamicclient /openv/netbackup/client/Solaris/Solaris2.5/bphdb /openv/netbackup/client/Solaris/Solaris2.5/bpjava-msvc /openv/netbackup/client/Solaris/Solaris2.5/bpjava-usvc /openv/netbackup/client/Solaris/Solaris2.5/bplist /openv/netbackup/client/Solaris/Solaris2.5/bpmount /openv/netbackup/client/Solaris/Solaris2.5/bprestore /openv/netbackup/client/Solaris/Solaris2.5/ftp_to_client /openv/netbackup/client/Solaris/Solaris2.5/install_client /openv/netbackup/client/Solaris/Solaris2.5/libvopie.so /openv/netbackup/client/Solaris/Solaris2.5/tar /openv/netbackup/client/Solaris/Solaris2.5/vauth_test /openv/netbackup/client/Solaris/Solaris2.5/vauth_testd /openv/netbackup/client/Solaris/Solaris2.5/vauth_util /openv/netbackup/client/Solaris/Solaris2.5/version /openv/netbackup/client/Solaris/Solaris2.5/vopie_util /openv/netbackup/client/Solaris/Solaris2.5/vopied /openv/netbackup/client/Solaris/Solaris2.5/xbp /openv/netbackup/client/Solaris/Solaris2.6/bp /openv/netbackup/client/Solaris/Solaris2.6/bpbackup /openv/netbackup/client/Solaris/Solaris2.6/bpbkar /openv/netbackup/client/Solaris/Solaris2.6/bpcd /openv/netbackup/client/Solaris/Solaris2.6/bpclimagelist /openv/netbackup/client/Solaris/Solaris2.6/bpclntcmd /openv/netbackup/client/Solaris/Solaris2.6/bpdynamicclient /openv/netbackup/client/Solaris/Solaris2.6/bphdb /openv/netbackup/client/Solaris/Solaris2.6/bpjava-msvc /openv/netbackup/client/Solaris/Solaris2.6/bpjava-usvc /openv/netbackup/client/Solaris/Solaris2.6/bplist /openv/netbackup/client/Solaris/Solaris2.6/bpmount /openv/netbackup/client/Solaris/Solaris2.6/bprestore /openv/netbackup/client/Solaris/Solaris2.6/ftp_to_client /openv/netbackup/client/Solaris/Solaris2.6/install_client /openv/netbackup/client/Solaris/Solaris2.6/libvopie.so /openv/netbackup/client/Solaris/Solaris2.6/tar /openv/netbackup/client/Solaris/Solaris2.6/vauth_test /openv/netbackup/client/Solaris/Solaris2.6/vauth_testd /openv/netbackup/client/Solaris/Solaris2.6/vauth_util /openv/netbackup/client/Solaris/Solaris2.6/version /openv/netbackup/client/Solaris/Solaris2.6/vopie_util /openv/netbackup/client/Solaris/Solaris2.6/vopied /openv/netbackup/client/Solaris/Solaris2.6/xbp /openv/netbackup/client/Solaris/Solaris_x86_2.6/bp /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpbackup /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpbkar /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpcd /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpclimagelist /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpclntcmd /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpdynamicclient /openv/netbackup/client/Solaris/Solaris_x86_2.6/bphdb /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpjava-msvc /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpjava-usvc /openv/netbackup/client/Solaris/Solaris_x86_2.6/bplist /openv/netbackup/client/Solaris/Solaris_x86_2.6/bpmount /openv/netbackup/client/Solaris/Solaris_x86_2.6/bprestore /openv/netbackup/client/Solaris/Solaris_x86_2.6/ftp_to_client /openv/netbackup/client/Solaris/Solaris_x86_2.6/install_client /openv/netbackup/client/Solaris/Solaris_x86_2.6/libvopie.so /openv/netbackup/client/Solaris/Solaris_x86_2.6/tar /openv/netbackup/client/Solaris/Solaris_x86_2.6/vauth_test /openv/netbackup/client/Solaris/Solaris_x86_2.6/vauth_testd /openv/netbackup/client/Solaris/Solaris_x86_2.6/vauth_util /openv/netbackup/client/Solaris/Solaris_x86_2.6/version /openv/netbackup/client/Solaris/Solaris_x86_2.6/vopie_util /openv/netbackup/client/Solaris/Solaris_x86_2.6/vopied /openv/netbackup/client/Solaris/Solaris_x86_2.6/xbp /openv/netbackup/db/bpenableLN.scr /openv/netbackup/db/bpenableTD.scr /openv/netbackup/dbext/install_dbext /openv/volmgr/bin/acsd /openv/volmgr/bin/acstest /openv/volmgr/bin/avrd /openv/volmgr/bin/goodies /openv/volmgr/bin/goodies/xvmadm /openv/volmgr/bin/ltid /openv/volmgr/bin/mmcrawl /openv/volmgr/bin/multihost_drive_notify /openv/volmgr/bin/oprd /openv/volmgr/bin/robtest /openv/volmgr/bin/sg.build /openv/volmgr/bin/sgscan /openv/volmgr/bin/tl4d /openv/volmgr/bin/tl4test /openv/volmgr/bin/tl8cd /openv/volmgr/bin/tl8d /openv/volmgr/bin/tl8test /openv/volmgr/bin/tldcd /openv/volmgr/bin/tldd /openv/volmgr/bin/tldtest /openv/volmgr/bin/tlhd /openv/volmgr/bin/tlmd /openv/volmgr/bin/tpclean /openv/volmgr/bin/ts8d /openv/volmgr/bin/ts8test /openv/volmgr/bin/tsdd /openv/volmgr/bin/tsdtest /openv/volmgr/bin/tshd /openv/volmgr/bin/tshtest /openv/volmgr/bin/vmcheckxxx /openv/volmgr/bin/vmconf /openv/volmgr/bin/vmd /openv/volmgr/bin/vmquery /openv/volmgr/database/external_robotics Problem Description: 4637467 nb clients are not updated w/ 108261-10 (from 108261-10) 4507040 potential data loss for media servers with multiple nics 4510169 NBU3.2J: bpsched schedules the same class two times. (from 112066-01) 4499273 September 9th, 2001 problem on jnbSA GUI. (from 108261-09) 4345788 Backup of a non exist directory exits with status 150 instead of 71 (from 108261-08) 4397701 Bpbackup -i (backup immediate) fails with error 25 (from 108261-07) 4394584 need Veritas NBU patch ...466 converted to patchadd format 4370437 some files would be unavailable for restore with exit code 1 (from 108261-06) 4341898 Cust ndmp associated drive being used by non-ndmp classes 4340667 Customer needs Veritas patches S0820343 and J0820412 in Solaris patchadd format (from 108261-05) 4323166 Interrupted system call & status 6 during Oracle OEBU backups. (from 108777-01) 4287760 Solaris 8 is not a selectable client for installation (from 108261-04) 4314679 updated need for patch 108261-03 to reflect Solaris 8 support (from 108261-03) 4303994 Veritas NBU 3.2 J0820326 patch, needs to be convert patchadd format 4292771 Customer needs patchJ0820226 from Veritas to fix his problem 4282898 Output of bpdbjobs cannot be formatted. (from 108261-02) 4286547 nbu getting 219 errors 4286545 bpexpdate returns an error when trying to change the expiration... (from 108261-01) 4244797 The usage information returned from sg.build utility is incorrect. 4267724 NBU 3.2 : possible corruption with full backups 4268538 Restores that need a large incremental take a long time to figure out what... 4268539 Provide support for robot serialization with the TLD robotic type 4268540 problems with bpretlevel 4268541 NBU 219 erroes associated with network time out and slave servers 4268542 When doing a TIR restore, the display of "is not in the true image list ... 4268543 MHD wizzard does work easily with only one tape drive 4268545 vmd shuts down due to too many file descriptors open 4268546 Add device support for the Compaq ESL9000 robotic tape library 4268547 problems with tape being stuck in drive; media can't dismount as expected 4268548 The release of NetBackup for Teradata Database Extension requires changes... 4268550 need improvements for Multihosted Drives (Shared Storage Option) Patch Installation Instructions: ------------------------------------------------------------------------- Refer to the Install.info file within the patch for instructions on using the generic 'installpatch' and 'backoutpatch' scripts provided with each patch. Any other special or non-generic installation instructions should be described below. ------------------------------------------------------------------------- Special Install Instructions: As root on your NetBackup Master Server: (for a NetBackup slave server, do steps 1-4) 1) Stop the NetBackup and Media Manager daemons: /usr/openv/netbackup/bin/goodies/bp.kill_all 2) Install patched binaries via patchadd/installpatch. 3) Run the new_clients scirpt /usr/openv/netbackup/bin/goodies/new_clients 4) Restart daemons. /usr/openv/netbackup/bin/initbprd /usr/openv/volmgr/bin/ltid -v 5) Update your NetBackup clients, with the update_clients script. /usr/openv/netbackup/bin/update_clients where is one of the following: Solaris Solaris2.5 Solaris Solaris2.6 Solaris Solaris7 Solaris Solaris8 Solaris Solaris_x86_2.6 Solaris Solaris_x86_7 Solaris Solaris_x86_8 Sun4 Solaris2.5 Sun4 Solaris2.6 Sun4 Solaris_dm Sun4 Solaris_dm_2.6 Remember to include the master server's type. Note: The /usr/openv/netbackup/bin/update_clients command without any parameters will update all the Unix clients. 6) If your master server is Solaris2.5 or Solaris2.6, For Solaris2.5: cp /usr/openv/lib/server/Solaris/Solaris2.5/libsfr.so /usr/openv/lib For Solaris2.6: cp /usr/openv/lib/server/Solaris/Solaris2.6/libsfr.so /usr/openv/lib Additional notes: =========================================== IV. UPGRADING 3.0 OR 3.1.1 TO 3.2 WARNING: =========================================== 1) Upgrade of 3.1.1 clients to 3.2 NetBackup client software fails. Install of new clients fails. When asked during an upgrade installation of NetBackup: "Do you want to update the NetBackup software on the clients?" Answer no and install the current NetBackup cumulative patch. Then execute /usr/openv/netbackup/bin/update_clients. See the description for patchP0820173 for complete details at the bottom of this README. 2) Links in /usr/openv/netbackup/client are not correct. This problem affects 3.1.1 to 3.2 upgrade of UNIX NetBackup master servers that push out their Netbackup client binaries, specifically the SGI or C910_920 client platforms. In 3.1.1: /usr/openv/netbackup/client/SGI is a link to /usr/openv/netbackup/client/C910_920. In 3.2: /usr/openv/netbackup/client/C910_920/IRIX6 is a link to /usr/openv/netbackup/client/SGI/IRIX6. The install for 3.2NetBackup attempts to relink existing directories to provide backwards compatibility. Both the Solaris pkgadd install and VERITAS ./install do a rm -f /usr/openv/netbackup/client/SGI to remove the link to C910_920. This remove has failed in some instances, resulting in a circular link of /usr/openv/netbackup/client/SGI/IRIX6 -> SGI. SOLUTION: Before upgrade to 3.2 ===================== Upgrading from 3.1.1 to 3.2, remove the directory, /usr/openv/netbackup/client/SGI prior to installing 3.2 NetBackup. rm -rf /usr/openv/netbackup/client/SGI Then proceed with installation of 3.2. ALREADY UPGRADED: =================== If the update to 3.2 has been completed: rm -rf /usr/openv/netbackup/client/SGI For Solaris servers, redo the pkgadd of SUNWnetbp, selecting SGI from the client list. For non-Solaris UNIX servers, select option 2 from the CDROM and install the SGI client binaries. The result should be the following links: SGI/IRIX5 -> C910_920/IRIX5 C910_920/IRIX6 -> SGI/IRIX6 (Note IRIX5 is not supported in 3.2GA so this link was done for 3.1.1 NetBackup clients.) There is no patch for this problem. 3) Entry for new client daemon is not correct in inetd.conf file The 3.2GA /usr/openv/netbackup/bin/install_bp creates an incorrect entry for vopied in the /etc/inetd.conf file, using /usr/openv/netbackup/bin instead of /usr/openv/bin . ( may be /etc/inet/inetd.conf or /usr/etc/inetd.conf for NCR) (may be /usr/etc/inetd.conf for SGI) When update_clients is run or a new client is installed, this incorrect vopied inetd.conf entry is propagated to the clients. If the server's inetd.conf is corrected before any update_clients or installs of new clients, the correct entry is made for vopied in the client's inted.conf file. All previously updated or installed clients need to have their inetd.conf entry for vopied corrected. edit the server's inetd.conf file (may be different for NCR and SGI) locate entry for vopied change /usr/openv/netbackup/bin to /usr/openv/bin. This will be corrected in upcoming releases. There will be no patch for this problem. 4) The official supported method to upgrade NetBackup is to install the new NetBackup binaries on top of the existing installed NetBackup binaries. The installation process recognizes the previous version of NetBackup and upgrades the existing binaries and in the upgrade to 3.2, performs a conversion of the Media Manager and NetBackup database files that includes a checksum. An alternative install method is to install the new version of NetBackup into a different directory or on to a different machine than the existing version of NetBackup. Then bprecover is run on the 3.1.1 (or 3.0) NetBackup's database files to the newly installed NetBackup 3.2 location. This alternate install method is not recommended. If the alternate method is used, work with a VERITAS Technical Support engineer and perform the following steps: 1. Backup the NetBackup 3.1.1 (or 3.0) database. 2. Install NetBackup 3.2. 3. Save the following files to another location. /usr/openv/volmgr/database/external_robotics /usr/openv/volmgr/database/external_mediatypes /usr/openv/volmgr/database/external_densities /usr/openv/volmgr/database/external_drivetypes /usr/openv/netbackup/db/Class_att_defs 4. With all NetBackup daemons stopped, run "bprecover" 5. Remove /usr/openv/volmgr/database/volDB 6. Copy the files saved in step 3 back to their original location. 7. Run /usr/openv/volmgr/bin/tpconfig -dbconvert 8. Start robotic and Media Manager daemons, /usr/openv/volmgr/bin/ltid -v 9. Check media manager GUI. Contact VERITAS support, if all tapes are not seen. 10. Check device manager GUI . Contact VERITAS support, if all tape drives and robots are not seen. 11. Start bpdbm and bprd, /usr/openv/netbackup/bin/initbprd If you have given nonroot users permission to perform NetBackup or Media Manager administration, (NetBackup System Administrator's Guide, Allowing Nonroot Users to Administer NetBackup page 31 and Media Manager Administrator's Guide, Allowing Nonroot Users to Administer Media Manager, page 9), you will have to reapply the group and file permissions changes as installing this patch will reset the group and file permissions to their delivered value. ******************************************************************** NOTE: this is only a Solaris server patch. Any other platforms are mention as clients. The rest of Veritas' readme is below. It contains references to other platforms. The following are descriptions of the problems fixed. Also included are 3.1.1 to 3.2 upgrade warnings. Please read the entire document before installing. Fixes with ** Description ** are important, they describe a problem that can lead to lost data. Please read these problem descriptions carefully. *********************************************************************** For NetBackup Java support, please check sunsolve.sun.com for the latest NetBackup Java patch for NBU 3.2 ================================== II. DESCRIPTION OF PROBLEMS FIXED ================================== The following are descriptions of the problems fixed. Please read the entire document before installing. Note: Any available workarounds to a problem are also listed. So another option would be to use the workaround instead of applying the patch. Fixes with ** Description ** are important, they describe a problem that can lead to lost data. Please read these problem descriptions carefully. If the NetBackup Java interface is used, obtain and install patch 112148-01. Patch 112148-01 must be installed with NetBackup patch 108261-06 or later. If patch 112148-01 was installed with a previous patch, do not re-install patch 112148-01 with this patch. See patch 112148-01 for further details. ============ Known Issues ============ Description: A media mount timeout can occur if the following conditions occur: 1)A restore ( or other non-backup process ) is started. 2)A backup is started attempting to use the same tape drive as the restore, but before the tape mount is complete. 3)If the backup starts after the mount has completed, the backup will queue in the scheduler. ============= Current patch ============= ================================================================================ Description: Backup jobs that fail with error 71, no files found, immediately after being started, may be incorrectly assigned exit status 150, termination requested. (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: To make partial restores easier, it is now possible to exclude items from a restore by adding a '!' character in front of a file or directory path specification. The exclude specifications may appear on the bprestore command or in the bprestore listfile. Once something is excluded, a subset of the files or directories can be re-included by a following include specification. Excludes can also be added when editing the selection list with the bpadm interface. Additional Notes: Restore excludes cannot be used when doing NDMP restores. (ALL NetBackup Servers and Clients) -------------------------------------------------------------------------------- Description: When restoring more than 25 AFS volumes at a time, the restore may fail. The progress log will contain a "can't open a pipe: Too many open files" message and TAR will exit with a status of 4. Workaround: Restore fewer AFS volumes at a time. (ALL NetBackup Servers, NetBackup Clients HP10.20/11.00, Solaris 2.5,2.6) -------------------------------------------------------------------------------- Description: In rare cases, duplication of images may hang. The bptm read and write processes complete, but the bpduplicate process does not terminate. (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: If there is a class defined: 1. With a Mac OS X Server client. 2. A file list of "ALL_LOCAL_DRIVES" 3. Streaming is not selected and the Mac OS X Server client has both UFS and HFS file systems, only the UFS file systems will be backed up. The HFS file systems will be skipped. Workaround: Define class with file list of "/" with the "cross mount point" attribute set to "yes". (NetBackup Server, MacOSXS) -------------------------------------------------------------------------------- Description: If the NetBackup server is setup to disallow client restore requests, the restore request would get denied but a restore job would still get created and appear in the 'queued' state until an administrator kills the job. (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: A BLIB backup with thousands of unchanged files can cause both the client and server processes to hang on a socket write to the other process. (NetBackup Clients HP10.20/11.00, Solaris2.5/2.6) -------------------------------------------------------------------------------- Description: When restoring directories with setgid bits, the sub-directories were restored with the parent directory's setgid bit setting instead of the sub-directory's original setgid bit setting. (ALL Unix NetBackup Clients) -------------------------------------------------------------------------------- Description: A denial of service attack, as reported at http://www.securityfocus.com/archive/1/175116, N+1 connections are made to the bpjava-msvc port (13722) via inetd, where N is the number of random ports available, can result in exhausting the number of ports available and a denial of service. The N+1st connection must be made within 1 minute of the first, as the first would timeout, creating an available port. Additional Notes: Do not open a 'port' on the master server for communication with a child user server when the master is transient (currently Unix only). This unnecesary port_acquisition can be the target of a denial of service attack. When transient, a new bpjava-msvc is created, via inetd, for every connection to 13722. It takes 60 seconds for bpjava-msvc to timeout if no valid request is sent following the connection, more than enough time to attempt multiple bpjava-msvc connections, and consume multiple (all) ports needlessly for communication with a user server which will never be created. Even if the transient master did receive a valid request and was able to create a (valid) user server, the information sent back to the master is only useful if the master is NOT transient, and is useless when the master IS transient. (ALL Unix NetBackup Servers and Unix NetBackup Clients) -------------------------------------------------------------------------------- Description: If there is more than one storage unit defined for a media server, and the specified hostname for each storage unit represents a different network interface (that is, the media server has more than one Network Interface Card), in a rare circumstance NetBackup may incorrectly deassign tapes, which could possibly result in data loss. Periodically, NetBackup starts a process that checks the media database against the image database, and expires media that no longer contain valid images. As part of this process, NetBackup queries the storage unit host for each storage unit to determine if it has a unique system name. The problem is that the system name string returned by this query was not null-terminated. If the order of the queries was such that a longer length system name was returned in between queries for the same host, the latter system name would be incorrect. For example, there are two media servers, one with two network interfaces. The hostnames specified in the storage units for the media server with the two network interfaces are "cat-nic1" and "ecat-nic2", and the system name returned for each should be "cat". The storage unit hostname for the other media server is "def" and the system name returned is "horse". Since the system name returned for the media servers is not null-terminated, the following shows the returned system names: storage unit host: cat-nic1 system name: cat storage unit host: def system name: horse storage unit host: ecat-nic2 system name: catse Since the returned system name is not correct for host cat in the last query, NetBackup assumes these are different hosts which leads to media being incorrectly deassigned. This problem is dependent on the order of the queries; if the storage unit host cat-nic2 was queried immediately after cat-nic1, as in the following example, the problem would not occur. storage unit host: cat-nic1 system name: cat storage unit host: cat-nic2 system name: cat storage unit host: def system name: horse Workaround: On Unix systems: Create the file /usr/openv/netbackup/bin/bpsched.d/CHECK_EXPIRED_MEDIA_INTERVAL on the master server and enter a 0 (zero) on the first and only line of the file. After this file has been created, bounce the NetBackup daemons. On an NT server: Create the file \NetBackup\bin\bpsched.d\CHECK_EXPIRED_MEDIA_INTERVAL on the master server and enter a 0 (zero) on the first and only line of the file. After this file has been created, bounce the NetBackup daemons After the workaround is in place, and some backups have occurred, run the bpexpdate command manually: bpexpdate -deassignempty -verbose If the command shows some media to be deassigned, it will query the user first. Answer 'n'. After generating a list of media: run "/usr/openv/netbackup/bin/admincmd/bpimmedia -mediaid " If an image is still assigned to that media id, it will return something like the example below: IMAGE canuck 3 canuck_0993794967 JIMTEST 0 testfull 0 7 2125 1017900567 0 0 FRAG 1 1 580576 0 2 13 6 CAN001 canuck 32768 201274 992536676 1 0 *NULL* 017900567 0 (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: The ascd process would not start with existing limit of drives in a STK LibraryStation configuration. The limit of drives in a STK LibraryStation configuration was increased from 940 to 1680 drives. The external symptom to this problem is ascd not starting. The system log file also contains the following error message: Apr 5 16:23:35 bison acsd[17118]: DecodeQuery() Actual status: Process killed by signal Apr 5 16:23:35 bison acsd[17118]: ACS(1) unavailable: initialization failed: Process killed by signal (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: The image files in the NetBackup image database do not get removed when Oracle images are expired through RMAN. (ALL NetBackup Servers) -------------------------------------------------------------------------------- Description: TapeAlert support enabled. (NetBackup Server, AIX RS6000) -------------------------------------------------------------------------------- Description: Release of the NetBackup Configuration Validation Utility (NCVU). (NetBackup Servers HP700/800, AIX RS6000, Solaris and Clients HP10.20/11.00 Solaris2.5/2.6) -------------------------------------------------------------------------------- Description: When browsing files or directories in the "Backup Files" tab of the NetBackup Java client GUI, any file or directory created after September 9, 2001 10:46:SS, will show a date that is in 1973. This behavior will also be seen in the "Task Progress" tab for any backup, archive or restore which was initiated after September 9, 2001 10:46:SS. Although the dates shown on the "Backup Files" tab and the "Task Progress" tab will be incorrect in the cases stated above, it will NOT prevent user backups, archives or restores. This problem does NOT affect the "Restore Files" tab. The dates on the "Restore Files" tab will be the actual date and time the files or directories were last modified and backed up. (ALL UNIX NetBackup Clients) ================================================================================ ============= patchJ0820495 ============= Description: Restores of files would change undefined group and user ids to the id of the account performing the restore. This is the correct behavior for non-root restores but for restores done by root the group and user ids should not be changed. (All NetBackup clients) Description: The bprd process timeouts may occur on client-initiated backups on random clients. The NetBackup server uses a timeout value of 10 seconds for connection requests. In some cases, due to a slow network, the number of hops, or some other reason, the connection may not be established in this time. This sort of failure results in the following messages in the bprd debug log: 07:52:12 [204] <16> db_ERROR: db_begin() failed: cannot connect on socket 07:52:12 [204] <16> logERROR: db_ERROR() failed: cannot connect on socket (25) bprd now looks for the file /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT. If this file exists, bprd uses the value contained in this file as the number of seconds for the connection timeout interval. For instance, if /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT contains the line: 20 then the bprd process times out connection requests after 20 seconds. If /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT does not exist, the bprd process uses a default timeout interval of 300 seconds. This has been increased from the previous default timeout interval, 10 seconds. bprd writes messages to its debug log to display the timeout interval, for instance: 13:22:47 [24038] <4> bprd: Looking for request-timeout file /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT 13:22:47 [24038] <4> bprd: File /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT found, resetting the request-timeout value to 20 seconds If the file /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT does not exist, the debug log contains messages like this: 13:18:03 [23964] <4> bprd: Looking for request-timeout file /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT 13:18:03 [23964] <4> bprd: File /usr/openv/netbackup/BPRD_REQUEST_TIMEOUT not found, using the default request-timeout value of 300 seconds (All NetBackup servers) Description: When doing block incremental restores with archive log images on the same tapes as the incrementals, the restore can hang if it has to wait to restore from the incrementals and starts restoring the archive logs which don't have any dependencies on previous block incremental backups. Workaround: Do a separate restore for the archive logs in a separate restore from the block incremental database files. (All NetBackup servers and clients) Description: 1) When using bprestore to perform an alternate path/alternate client restore from one NT client to another NT client, the alternate path substitution may cause the restore to fail. The following error message is displayed in the logs: (WIN32 123: The filename, directory name, or volume label syntax is incorrect. ). This is due to an extra \r character being written to the rename file when it is written on the destination client. 2) When performing an archive operation from a NT client and the NT client is setup to use OTM, the archive log always reports that the OTM license has not been installed on the server regardless of it being installed or not. Workaround: 1) Create the rename file without the original \r character. Most text editors on NT will automatically add this character to the file. Use textpad to allow the rename file to not contain the \r character. 2) There is no workaround to this problem. (All NetBackup servers) Description: Check the last update time of a volume against the "last backup time". If the last update time of a volume is later than the last backup, then "vos backup" and "vos dump" commands are issued from the AFS portion of bpbkar. Otherwise the backup and dump commands are skipped, and bpbkar proceeds with the next AFS volume. This can drastically reduce the amount of time a backup would take, given many AFS volumes. If the "verbose" option is set in the NetBackup config file, skipped files will be logged in the bpbkar log file. (All UNIX NetBackup clients) Description: Backup '.' files were allowed when file paths of the form .../* are used. This caused mount points at the level .../* to not be crossed if the cross mount points option is off. The method of handling the '.' files was changed so that the first level mount point will always be crossed when specified with a /*. Workaround: Instead of specifying paths using .../* use .../*.* (All NetBackup clients) Description: During restores, sparse files can get I/O errors on the ftruncate call on Sequent 4.2 and later. The error message in the progress log is: Couldn't truncate file /abc to 10000 bytes: I/O error (NetBackup clients, Sequent4.2) Description: When doing software compression on an already compressed file that is over 2 gigabytes long, the size information is not set properly during backups which causes the restore to fail with the following message: invalid file size encountered for file /data/abc.Z (All NetBackup UNIX clients) Description: A verify or an import of a backup image that has a path name that contains a newline character (\n) and the path length is greater than 100 characters fails. (All NetBackup UNIX servers and clients) Description: Backups that need the final server status can fail if the socket that the server status is sent on does not remain active so that the status can be read. (All NetBackup UNIX clients) Description: If the same image is needed by more than 1 restore at the same time, one of the images can fail with: 16:51:07 (9265.007) Status of restore from image created Tue Oct 03 07:31:48 2000 = socket write failed (All NetBackup servers) Description: When dealing with a large number of files, a failure can occur with the message "Timed out after X seconds waiting for the file list" in the bpbkar log. The original problem was noted (with VSM) when running mignospace against 10,000 files in a single directory. The original value used for the timeout (FILELIST_TIMEOUT) was 60 seconds. This was increased to 300 seconds. If the file /usr/openv/netbackup/FILELIST_READ_TIMEOUT exists, a value (in seconds) will be read from it. If this value is greater than or equal to 60, it overrides the default. (All NetBackup UNIX clients) Description: The "CRYPT_OPTION = required" configuration entry on a server, will cause a catalog database backup (bpbackupdb) on that server to fail with an error 51. Workaround: Use "CRYPT_OPTION = allowed" instead. Additional Notes: Catalog database backups will not be encrypted even if "CRYPT_OPTION = required" is specified. (All NetBackup servers) Description: If OTM is licensed on a server running NetBackup 3.4 and an archive is attempted from a UNIX client running NetBackup 3.2, the files will be backed up but not deleted. An error status 3 will be given along with a similar message: valid archive image produced, but no files deleted due to non-fatal problems. Workaround: Delete the files by hand after the archive is complete. (All NetBackup clients) Description: If duplicating SQL images that were not a multiple of 1024 bytes, the write side of bpduplicate would fail. The check for ensuring images are multiples of 1024 bytes has been modified to ensure its a multiple of 512 bytes. Not all of the shared memory used by the bptm process may be released following a duplication. This problem only occurs when using multiplexed duplication. Workaround: Use the ipcrm -m command to free the orphaned shared memory. (All NetBackup servers) Description: A mount request to a non-zero LUN robotic drive on AIX will fail with the drive being downed. The following is a log of what happens when a non-zero LUN robotic drive fails when trying to unmount a tape. Messages file with VERBOSE setting in vm.conf file: Sep 18 16:02:46 slbdev1 tldd[25960]: TLD(0) DismountTape 000017 from drive 2 Sep 18 16:03:02 slbdev1 tldd[22286]: TLD(0) unload==TRUE, but no unload, drive 2 (device 1) Sep 18 16:03:03 slbdev1 tldd[25960]: DecodeDismount(): TLD(0) drive 2, Actual status: Robotic dismount failure Sep 18 16:03:03 slbdev1 tldd[25960]: TLD(0) drive 2 (device 1) is being DOWNED, status: Robotic dismount failure Sep 18 16:03:03 slbdev1 tldd[25960]: Check integrity of the drive, drive path, and media Workaround: Use drives on LUN zero only. (NetBackup servers, RS6000) Description: An SSO installation with a large number of media servers could experience scheduling slow downs when several backups are executed using specific storage units. Queries were made to all the media servers, causing scheduling interruptions that lasted several minutes. Communications are now asynchronous so the scheduling process does not wait for completion. (All NetBackup servers) Description: The vmd process (Media Manager volume daemon, NetBackup Volume Manager service) on the DA (device allocation) host core dumps in an SSO environment with more than 15 media servers. If the /volmgr/debug/daemon log file was enabled, the last lines written to the log file would be similar to the following: 09:42:07 [9090] <2> da_register_shared_drives: server read: 5 1 - drive0 drive1 drive2 drive3 drive4 09:42:07 [9090] <2> da_register_shared_drives: returning scan host record for 5 drives registered by host .any.where.com for host Workaround: Limit the number of SSO media servers to 15 or less. Use the MH_SCAN_ABILITY configuration setting and ensure that the scan host is one of the first 15 servers that has registered its drives with vmd/DA, the device allocator. (All NetBackup servers) Description: When a broken volume or mis-configured cleaning tape is encountered using a TL8 robot, the result is all tape drives are downed in the robot. (All NetBackup servers) Description: STK 9840 and T9940A drives do not support the CLEAN NOW TapeAlert flag. Changes were made to the TapeAlert drive cleaning mechanism to check for a flag, CLEAN PERIODIC, in addition to the CLEAN NOW flag. This new flag will be checked for all drive/platform combinations that support TapeAlert. (All NetBackup servers) ============= patchJ0820466 ============= Description: Attempts to start a manual or immediate backup of an incremental schedule with multiple data streams enabled will fail with a 198 error. In addition, incremental backups using multiple data streams may be initiated every time the bpsched process is started. This problem was introduced by patch J0820465. (All NetBackup servers) NOTE: Client patches have not changed from patchJ0820465 and will continue to display this version. ============= patchJ0820465 ============= Description: There is a message queue used by main bpsched process for communicating storage unit information. This queue is created when a reread storage unit is initiated automatically by bpsched. It is removed by bpsched when the information has been received. If the main sched process should terminate abnormally via kill -9 or some type of failure, this queue will not be removed properly. This will cause the automatic reread storage units queries to fail until the message queue is removed or the system is rebooted. This patch will remove this message queue automatically when main bpsched process is initiated. Workaround: Manually removing the message queue identified by the key 0x52455245. This can be done with the ipcrm command. The queue id needed for the ipcrm command will be displayed by a ipcs -q. (All NetBackup servers) Description: Currently, the NetBackup Java GUI (jbp) "logs into" a NetBackup UNIX client (bpjava-msvc) using a UNIX user name and password. bpjava-msvc uses the getpwnam(3) (and possibly getspnam(3)) to authenticate the user name and password. Some UNIX platforms, like Solaris 2.6 and later, support Pluggable Authentication Modules (PAM). This makes it possible to support other authentication methods like Kerberos semi-transparently on the system. Applications can be made PAM-aware by calling functions like pam_authenticate(3) and letting the PAM libraries figure out how to use the configured authentication method. The 3.2 GA version of bpjava-msvc is not PAM-aware. If the system is configured to use Kerberos authentication, bpjava-msvc will not be able authenticate a log in attempt because the password value in /etc/shadow is not meaningful. This patch makes bpjava-msvc PAM-aware. bpjava-msvc will attempt to use PAM authentication first to validate a user name/password. If that fails, it will use the traditional UNIX shadow password. Additional Notes: Currently, this patch is only available for the Solaris 2.6 and Solaris 2.6 X86 clients. By default, bpjava-msvc uses the "login" PAM service as defined in the pam.conf(4) file to do its authentication. To use an alternate PAM service, create this file: /usr/openv/netbackup/pam_service.txt The first line of the pam_service.txt file should be the alternate PAM service name. (NetBackup Solaris clients) Description: The "avrd" process consumes much more CPU time than expected in some SAN media server configurations. Under heavy mount/dismount activity, if there are no locally attached drives configured on a media server, and if the media server is not serving as the scan host for any multihosted drives, the avrd process uses excess CPU time. Workaround: Configure a local drive on SAN media servers for avrd to scan. (All Media Manager servers) Description: In the bpbrm log on the NetBackup server when either bpbkar or tar is initiated, the following error message will be encountered: ld.so.1: fatal: relocation error: file /opt/lotus/notes/latest/sunspa/libnotes.so: symbol verifyAtomics: referenced symbol not found (NetBackup Solaris 2.6 clients) Description: If additional records are added to a Lotus Notes database while the database is being backed up, the backup may be corrupt. Restore of a Lotus Notes database greater than 2 GB in size will fail. The import or verify of a Lotus Notes image that contains a Lotus Notes database greater than 2 GB in size will fail. (NetBackup Solaris 2.6 clients) Description: If there are a large number of disk storage units, and a large number of disk backups are queued that can run quickly, the scheduler will get tied up processing job complete messages and starting new disk jobs. While this happens, the request queue is not serviced. This causes the bpsched processes that are building worklists to wait. The result is that tape drives will remain idle, when work should be initiated on them. Workaround: Set the schedule windows of the tape backups to open before the disk backup jobs. This causes them to be loaded on the scheduler work queue before the disk backups cause the condition. (All NetBackup servers) Description: 1) If a user directed backup is initiated with a progress log, and the master is blocked from client access of the progress log by a firewall, the main bpsched process will hang for client_connect_timeout seconds. During this time, no jobs will start. This condition is now blocked by bprd attempting to update the progress log before starting the jobs. The job is failed with an entry in the problems log if the progress log update fails. 2) Extra data streams will be generated if: a class is created with 1 or more schedules, multiple date streams is enabled, a backup is initiated, a schedule type is changed, then another backup is started. The problem was that the streams file ends up with 2 schedules of the same name, with different types. The scheduler would then start all the streams of the same name even though the schedule types were different. 3) The Scheduler had a problem handling interrupted system calls correctly when the message queues were full. If a signal occurred while a scheduler process was waiting to write a message on the message queue, the process would exit with an error status. This has been corrected so that the write is attempted again if interrupted system call is detected. 4) The scheduler in Patch J09820443 does not support media servers with older patch levels if the server has multihosted drives. If scheduler communicated with an older rev media server, scheduler would crash and the reason was difficult to determine. This patch detects the problem, logs an error in the problems log specifying the server of incorrect revision level, and disables multihosted drives for that storage unit. (All NetBackup servers) Description: An import of an image backed up with Multiple Data Streams will fail if imported on another host that does not have Multiple Data Streams enabled for the client. The import receives a "file open failed" error (code 12), but the import does actually import the image which can be successfully used for restores. When this problem occurs, the following message is seen in the import progress log: 09:58:29 INF - Import of class class_name, schedule full (client_name_0967215125) failed, file open failed. The bpdbm log will contain entries similar to the following: 09:58:29 [2078] <16> part_validate: cannot open STREAMS file for client 09:58:29 [2078] <16> validate_image: failure to update STREAMS for client client_name: file open failed (12) 09:58:29 [2078] <4> bpdbm: request complete: exit status 12 file open failed Workaround: The import actually is successful. Restores can be done from the imported image, so the error status can be ignored. (All NetBackup servers) Description: For HP800, provide application error recovery for backups/restores. This error recovery will attempt to recover from intermittent fibre channel conditions that have nothing to do with the media itself. Also, for all platforms that support application error recovery, correct problem where the KBS/sec message for non-MPX backups was not accurate if an error recovery occurred. (NetBackup servers, Alpha, Windows NT Alpha/Intel, HP800 and Solaris) Additional Notes: For application error recovery to be work on HP800, there must be access to the tape drive via a passthru driver. Refer to the NetBackup Media Manager Device Configuration Guide for information how to create device files for passthru for tape drives. Description: When backing up NT NetBackup clients and backup notify scripts are being used, there is a chance that the sockets may timeout from under NetBackup if the notify script takes longer than 3 minutes to execute. (All NetBackup servers and NetBackup Windows NT Alpha/Intel client) Description: When doing incremental backups on directories with many 1000s of files, the default CLIENT_READ_TIMEOUT is not large enough to prevent timeouts when there are very few files that have changed. Workaround: Use a larger CLIENT_READ_TIMEOUT than the default of 5 minutes. (All UNIX NetBackup clients) Description: During large restores, bprd can get an allocation failure when it gets up to 256MB in size. Workaround: Do multiple restores with fewer files in them. (NetBackup RS6000 clients) ============= patchJ0820443 ============= Description: Files with uids and pids greater than 18 bits in length were restored with incorrect uids and pids. (NetBackup client, RS6000) Description: 3 different messages had the same I18N message code. Changed the messages about Couldn't mkdir /usr/openv/bin and /usr/openv/var/auth from CerIns0230N to CerIns0231N, which has the directory as a variable instead of a part of the translated message string. (All NetBackup servers) Description: When running robot inventory update with a TLM robot selected, if more than 1000 volumes are in the robotic library, the update fails with Media Manager error 34, "Media ID not unique in database". Workaround: Run the robot inventory "Show Contents" report, then add the new volume entries with "Actions, New, Volumes" (Add New Volumes dialog) in the Media and Device Management interface. Use the move and delete operations when media has been taken out of the robotic library. (All Media Manager servers) Description: A restore fails with a 174 error (media manager system error occurred). The bptm log displays the following: > 15:43:11 [31134] <2> wait_for_sigcld: waiting for child to exit, timeout is > 10000 > 15:43:11 [31134] <2> send_brm_msg: ERROR 174 > 15:43:11 [31134] <2> mpx_terminate_exit: EXITING with status 174 (All NetBackup servers) Description: When using NDMP with a library that has drives split between two or more NDMP servers, a backup on one server may try to reuse an already mounted media on another server even though a local drive is available. A different media will now be selected and mounted into the local drive. (NetBackup servers, HP700/800, RS6000, Solaris, Windows Alpha/Intel) Description: Early phases of NDMP backups may timeout for large volumes after 6 hours with a bptm log message of "aborting backup - no mover progress". The default has been increased to 8 hours and is now configurable via the .../netbackup/db/config/NDMP_PROGRESS_TIMEOUT file. The file format is a single line containing the number of minutes. The value must be greater than 10 minutes and less than 1440 minutes (24 hours). (NetBackup servers, HP700/800, RS6000, Solaris, Windows Alpha/Intel) Description: NDMP restores to alternate paths of names containing spaces did not work. (NetBackup servers, HP700/800, RS6000, Solaris, Windows Alpha/Intel) Description: A second NDMP restore that requires a media that is in use can cause bprd to timeout while waiting for the media. Once the media becomes available the restore failed. (NetBackup servers, HP700/800, RS6000, Solaris, Windows Alpha/Intel) Description: The hard link lookup takes too long with a large number of hard links. (All UNIX NetBackup clients) Description: A catalog backup fails with the following error message: "NB database backup failed, a path was not found or is inaccessible." The admin log shows an entry similar to the following (it contains the Lff or LFF code): 09:14:44 [6958] <16> readline: from host server_name, Lff - 1073741823 1 0 1 1 3 40108039 21 33188 root other 1073741823 958161403 958163889 958178867 /usr/openv/netbackup/db/images/server_name/0955000000/test_0955523188_FULL.f The file (.f) that is being backed up is approximately 1-2 GB in size. (All NetBackup servers) Description: NDMP temporary inomap files can be left after a backup terminates due to an error. (NetBackup server, HP700/800, RS6000, Solaris, Windows Alpha/Intel) Description: When doing NFS backups of Network Appliance boxes on Solaris, if async I/O is used on a system with heavy load, the system sometimes returns EINVAL on the aio_read() call. When async I/O is disabled by creating file /usr/openv/netbackup/AIO_READS_MAX with a value of 1 in the file, the sparse file handling code still attempted to do async reads. Code was added to the sparse file handling code to do normal reads if the async I/O is disabled. The bpbkar log will have the following message when the problem occurs: 20:28:46 [383] <16> bpbkar dump_sparse: ERR - Read error at byte 233462 reading 1024 bytes in file /data/object/pbrboot.DB. Errno = 22: Invalid argument (NetBackup client, Solaris2.5, Solaris2.6) Description: The following error is seen when attempting to restore files: "Attempting to restore migrated files on a system with no HSM installed." However the files are NOT migrated and there is NO HSM installed on system. Many, many files restore correctly before the failure occurs. (NetBackup client, Solaris2.5, Solaris2.6) Description: 0434, Using bpbkar in patch J0820412, if a directory contained both migrated files and other subdirectories, those subdirectories were incorrectly dumped as migrated files. Here is an example. This problem was observed on Solaris and Irix systems and may have occurred on HP-UX systems. This problem only occurs on filesystems managed by VERITAS Storage Migrator. (NetBackup clients, HP9000-800/HP-UX11.00, HP-UX10.20, SGI) Description: The temporary memory used by bpsched to hold the file list for user backups was not being freed when the backup finished. If there are many user directed backups with very large file lists for the backup, this becomes a problem. Bpsched main grows in size, and the process it forks will start with more memory. Eventually, the system runs out of memory. The result are problems that appear to be system problems. This patch corrects the problem by freeing the memory when user directed backups finish. (All NetBackup servers ) Description: Robot inventory of an LMF, TLH, or TLM robot fails due to "invalid media type" when an unsupported vendor media type is encountered. An extended error message such as "Unrecognized TLH media type UNKNOWN" may be returned. Workaround: Use only vendor media types that are officially supported, refrain from using robotic inventory to manage media, or use the New/Change/Move/Delete Volume commands to manage media. Additional Notes: This situation occurs in NetBackup 3.2 when IBM 3590 K (extended length) cartridges are encountered in TLH robots. (All Media Manager servers) Description: When the Media Manager volume daemon (UNIX) or NetBackup Volume Manager service (NT) (vmd) requests a volume for use, the maximum mounts parameter of the volume isn't correctly checked. This was allowing the tape requester to request a tape that had already exceeded the maximum mounts volume parameter. Workaround: For a volume that has reached maximum mounts, the volume can either be removed from the volume pool or frozen. Or set maxmounts to infinite (zero). (All Media Manager servers ) Description: bpexpdate -recalculate -class -ret -sched changes the retention level for all backups done to the specified class. The -sched option is not used to further qualify the set of backups on which the recalculation is done. For example, bpexpdate does not support changing the retention period on just the incremental backups and leaving alone the retention level on the full backups. Workaround: Specify to bpexpdate the individual backups for which the retention needs recalculation. The -backupid option allows an individual backup to be specified. For instance, run bpexpdate -recalculate -d date_time -backupid backup_id (All NetBackup servers) Description: On HP 11.00 systems, EINVAL was returned on the async_read64 call on the 100th call to async_read64. The logs would show error messages like: 12:57:57 ERR - Read error at byte 6479360 reading 65024 bytes in file /ORACLE/arch/WTCR0022/arch.log1_1123.dbf.Z. Errno = -1: Workaround: Create file /usr/openv/netbackup/AIO_READS_MAX with a 1 in it on the client. ( NetBackup client HP9000-800/HP-UX11.00 ) Description: Different NDMP server platforms report media errors in different ways. NetBackup did not handle a NDMP_NOTIFY_PAUSED with a reason code of NDMP_MOVER_PAUSE_MEDIA_ERROR. (NetBackup servers Windows Alpha/Intel, HP700/800, RS6000, Solaris) Description: Bpduplicate fails when duplicating an image from a tape that is in use by another bpduplicate process or is in use by a restore. The bpduplicate fails due to a media open error (error 83). The bpduplicate logs indicate the following error: 14:42:08 INF - Media id B00004 on server server_name is in use, waiting for it. 14:42:30 INF - Beginning duplication on server server_name of client client_name image, creating copy 2. 14:52:12 INF - Waiting for positioning of media id B00004 on server server_name. 14:52:58 INF - host server_name backup id client_name_0961780631 read failed, media open error (83). The bptm logs indicate that another process did a tpunmount of the media-id requested by the bpduplicate process, which causes the open error: 14:52:15 [16782] <2> tpunmount: tpunmount'ing /usr/openv/netbackup/db/media/tpreq/B00004 14:52:15 [16782] <2> getsockconnected: host=server_name service=bpdbm address=166.98.24.43 protocol=tcp non-reserved port=13721 14:52:15 [16782] <4> mpx_read_backup: successfully restored 1 of 1 requests, read total of 204084 Kbytes at 324.483 Kbytes/sec 14:52:15 [16782] <2> bptm: EXITING with status 0 <---------- 14:52:24 [16873] <2> io_position_for_read: locate block is done 14:52:24 [16873] <2> getsockconnected: host=server_name service=bpdbm address=166.98.24.43 protocol=tcp non-reserved port=13721 14:52:24 [16873] <16> io_open: cannot open file /usr/openv/netbackup/db/media/tpreq/B00004, No such file or directory 14:52:24 [16873] <2> log_media_error: successfully wrote to error file - 06/27/00 14:52:24 B00004 0 OPEN_ERROR Workaround: This problem can be avoided by not having more than one bpduplicate process simultaneously duplicating from the same media-id. Also, do not have a bpduplicate process simultaneously duplicating from a media-id that is being used by a restore. Additional Notes: Added the following I18N messages: SmeBtm2650N "successfully duplicated backup id %s, copy %d, %d kbytes" SmeCom2615N "successfully wrote %d of %d multiplexed duplications, total Kbytes %d at %.3f Kbytes/sec" (All NetBackup servers) ============= patchJ0820413 ============= THIS PATCH MUST BE INSTALLED BY ALL STORAGE MIGRATOR USERS. NETBACKUP ONLY CUSTOMERS WITH THE LATEST CUMULATIVE PATCH DO NOT NEED TO INSTALL THIS PATCH It has been discovered that versions of NetBackup are capable of generating backup images for file systems managed by VERITAS Storage Migrator that will identify directory structures as files. When doing a full or disaster recovery restore, messages similar to this will be seen in the TAR restore log 11:33:04 (35.001) /hsmt1/archivet1/druid2/ 11:33:04 (35.001) /hsmt1/archivet1/druid2/filet1 11:33:04 (35.001) Could not create migrated file /hsmt1/archivet1/druid2/filet1: : Not a directory 11:33:04 (35.001) /hsmt1/archivet1/druid2/filet10 11:33:04 (35.001) Could not create migrated file /hsmt1/archivet1/druid2/filet10: : Not a directory 11:33:04 (35.001) /hsmt1/archivet1/druid2/filet11 These messages are produced because a file, /hsmt1/archivet1/druid2, is created preventing the files that belong in the directory /hsmt1/archivet1/druid2/ from being created. Individual and group file restores are still possible; as are restores into an existing directory structure. This problem is only known to affect directories that contain files actively managed by Storage Migrator. Workaround: If the above problem is experienced, contact VERITAS Customer Support (Storage Migrator support group) to assist in the restoration of files. VERITAS Customer Support will require a 'tar -tvf' of the backup image file producing the above problem to assist in the full restoration of the file system. This will assist in the creation of the directory structure. Once the directory structure is rebuilt (via mkdir -p /manage_filesystem/subdir/subsubdir), true files can be restored over the existing directory structure utilizing the 'tar xPf image_file' command. This patch will alleviate the above problem. ============= patchJ0820412 ============= Description: The message queues used by bpsched fill up, causing bpsched to hang. This problem was corrected on most platforms 2 years ago, but HP OS levels would not support the fix. HP version 10.20 allows the fix to work. This change is simply a compile change to include the changes in hp700 and hp800 platforms. (NetBackup hp700, hp800 servers) Description: xbp could crash with a segmentation violation if the directory in the "Directory Structure" list was doubled-clicked with the mouse and then the "Edit" menu item "Select all in current directory" was selected. (All UNIX NetBackup clients) Description: When upgrading from 3.1.1 NetBackup to 3.2 NetBackup, update_clients fails because it can't find the authentication directories, /usr/openv/bin and /usr/openv/var/auth. The first solution was in a server binary, bpinst, worked for all clients except SCO/OpenServer and ALPHA/OSF1 3.2 NetBackup clients. A second solution (implemented in a client binary) works for updating clients after the first solution but not for an initial upgrade from 3.1.1 to 3.2GA NetBackup. This solution is implemented in a server binary and works for all client updates and all NetBackup client platforms. (All UNIX NetBackup servers and clients) Description: Bpduplicate, bpimport, or bpverify crash when there is a problem writing to the progress log. (All NetBackup servers) Description: Increased the number of slots supported by the TL8 robotic daemon. Increased the number of drives supported by the TL8 robotic daemon. This was done for the ADIC Scalar 100 with AIT drives. (All Media Manager servers) Description: Fixed the tpclean usage statement to reflect the use of drive names instead of device names. (All Media Managers servers) Description: When restoring to an alternate client without entries in the /etc/passwd and /etc/group file for the same user and group names as the backup machine, the files get restored using the UIDs and GIDs stored in the tar headers. When UIDs and GIDs are larger than 262143, they overflow the space provided for them in the tar headers using the current encoding scheme. Using all the space provided in the tar headers, however, can extend this limit up to 16777215. Workaround: Add entries to the /etc/passwd and /etc/group files so that the user and group names can be looked up. (All UNIX NetBackup clients) Description: 1) If a class has a required storage unit specified, and no drives are available because one or more drives are in use for non-backup purposes, the backups fail with a 219 status. Prior to patch J0820326, they would go active but queue in media manager until either the drives became available or a media mount timeout occurred. This change puts them into a queued state until the drives are available at which time the job goes active. 2) Backups normally run on whatever server has drive capacity available. This causes a problem, especially in Multihosted Drives environments, where backups that could be done locally, are sent across the network, reducing performance. This patch allows NetBackup to be configured to require backups be done locally if the client is also a server. This is done by adding the line MUST_USE_LOCAL_DRIVE in the bp.conf file, or in the NT registry depending on what the server is. 3) If no drives were available due to being down, an error 96 would occur indicating there is no media available. This has been corrected to return the correct status. 4) The last_backup time that was put in the bpsched debug log was printing the address of the date instead of the actual date. This has been corrected. 5) The process that does the automatic reread storage unit was not disassociating itself from the parent properly, causing the NetBackup scheduler to get the signal, SIGCLD, when the process terminated. This could confuse the NetBackup scheduler. Workaround: 1) Set QUEUE_ON_ERROR and WAIT_IN_QUEUE in the NetBackup server bp.conf file. For NT, set in the registry, as type DWORD with a value of 1, HKEY_LOCAL_MACHINE\SOFTWARE\VERITAS\NetBackup\CurrentVersion\Config With these entries, a job that attempts to start when this condition occurs will be requeued with a 219 status and wait for the storage unit to become available. (All NetBackup servers) Description: The bprd process runs out of memory space when trying to restore a single file if there are a large number of images for the client and the bp or xbp interface was used to select the file to restore. Workaround: Use a date range with the java GUI or use bprestore with a date range. (All NetBackup servers) Description: The UNIX message queue can get filled with messages that will never be read. The main sched would send a message to the child process, it would check if the child was still alive. If the child timed out waiting for main sched, the child may end up in a defunct state and the OS would report to main sched that the process was still alive. Some older HP OS levels did not support the changes to correct this. Current HP OS levels do support the changes and this problem should no longer happen. Workaround: There is an executable that can be used to periodically clean the message queues to prevent them from becoming full. Additional Notes: This patch must be installed after patchJ0820363 (NetBackup hp700, hp800 servers) Description: At the end of a FlashBackup, bptm reads the .f file from disk and writes it to the tape. There was a race condition where bptm could read the file before bpbrm had finished writing it to disk. bpbrm now flushes any .f information to disk when it sees the raw partition entry added by bpbkar. (All NetBackup servers) Description: 1) Prior modifications to the master config file initialization broke the busy files feature. The busy_file_directory field was set to default NULL, instead of defaulting to: /usr/openv/netbackup/busy_files. This resulted in the following error message when the busy file feature is enabled: "Skipping busy file processing, error in master configuration file" 2) The busy files feature reported files greater than 2 gigabytes were busy, when they were not. Workaround: For the first problem, add the following entry to the bp.conf file: BUSY_FILE_DIRECTORY = /usr/openv/netbackup (All UNIX NetBackup clients) Description: If the user's bp.conf file did not specify a BUSY_FILE_DIRECTORY, the busy files feature generates the following error message: "Skipping busy file processing, error in user's configuration file" This is incorrect. The feature should use the user's home directory as a default if this option is not specified. Workaround: Add the following busy files option to the user's bp.conf file: BUSY_FILE_DIRECTORY = $HOME (All UNIX NetBackup clients) Description: Correct problem where the *.t file in /usr/openv/netbackup/db/jobs did not always show the correct transfer rate in the KBW entries. This problem occurred on MPX backups. (All NetBackup servers) Description: If the NetBackup image database and media database are inconsistent, media may be incorrectly deassigned. This situation may be caused by the incorrect use of the bpimage -newserver -oldserver command, where the new server is actually an existing media server. Since this command updates the image database but not the media database, inconsistencies may lead to media being deassigned. The All Log Entries report would show entries similar to the following: media id A00000 removed from media manager database (manual deassign) Deassigned media-id A00000 on host s2 media id B00000 removed from media manager database (manual deassign) Deassigned media-id B00000 on host s2 Deassigned 2 out of 2 media-ids found. The Images on Media report would show that images are still present for one or more of the expired media on host s1. Workaround: On UNIX systems, create the file /usr/openv/netbackup/bin/bpsched.d/CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. On an NT server, create the file \NetBackup\bin\bpsched.d\CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. This will cause NetBackup to behave as it did in the 3.1.1 release. (All NetBackup servers) Description: Add ACS support for NDMP. (Media Manager hp700, hp800, Windows NT Alpha and Intel, rs6000 and solaris servers) Description: A file, directory, or filesystem containing the "%" character (and possibly other special characters) may not be reported to bpcoverage or bpsched (for multiple data streams). In addition, bpmount on the client may crash. (All NetBackup clients) Description: Bpbkar was core dumping causing exit status 130. (NetBackup HP 11.00, Solaris2.5, Solaris2.6 clients) Description: patch J0820363 contained HP-UX11.00 client binaries, bpbkar and tar, compiled with a Storage Migrator 3.25GA libmigsmall.a. This provides the correct bpbkar and tar binaries for HP-UX11.00 HSM 3.2GA NetBackup clients and creates a new /usr/openv/netbackup/client/HP9000-800 directory for HSM3.25GA NetBackup clients with a bpbkar and tar only. The correct bpbkar and tar binary must be applied to HP-UX11.00 NetBackup clients running HSM3.2GA to prevent cached files. (NetBackup HP 11.00 clients running HSM 3.2GA or HSM 3.25GA) Additional Notes: Copy the appropriate bpbkar and tar to HP11.00 NetBackup clients running HSM 3.2GA and HSM 3.25GA. For HP11.00 NetBackup clients running HSM 3.2GA, echo "HP9000-800 HP-UX11.00 hsm32_client1" > /tmp/HSM32clients echo "HP9000-800 HP-UX11.00 hsm32_client2" >> /tmp/HSM32clients update_clients -ClientList /tmp/HSM32clients For HP-UX11.00 NetBackup clients running HSM 3.25GA, cd /usr/openv/netbackup/client/HP9000-800/HP-UX11.00_33 If HSM 3.25GA is running on the NetBackup master server, cp bpbkar /usr/openv/netbackup/bin cp tar /usr/openv/netbackup/bin If HSM 3.25GA is running on a NetBackup client, hsm325_client1, rcp bpbkar hsm325_client1:/usr/openv/netbackup/bin rcp tar hsm325_client1:/usr/openv/netbackup/bin Description: Modification to the application error recovery (first introduced in patchJ0820260) to allow it to recover fibre related errors on Solaris 2.7. (NetBackup solaris servers) Description: The process that bpsched uses to detect that unavailable drives have become available, failed on the ALPHA NetBackup server platform. This problem only occurs if all drives become unavailable due to being down, or in use for a non backup such as a restore or duplicate. When this happens, all jobs remain queued waiting for storage units that will never become available. Workaround: A workaround for this is to execute the following command: /use/openv/netbackup/bin/admincmd/bpschedreq -read_stunits This will instruct sched to update the drive status immediately. This can be setup as a cron job to run every 10 minutes if necessary. (All NetBackup servers) Description: NetBackup will attempt to backup filesystems of type procfs and cdrom if they are not skipped using an exclude_list. This will cause error messages similar to: ERR - Could not get link name of... in the NetBackup log. This fix will skip filesystems of type procfs. Workaround: Explicitly exclude the filesystems by putting path in a NetBackup exclude_list file. (NetBackup Linux clients) ============= patchJ0820363 ============= Description: Attempts at cleaning drives in a TLM (DAS-controlled) robot caused ltid, the Media Manager device daemon, to core dump. In the system log, messages such as the following may appear: "Daemon terminating because ltid is no longer active" Workaround: Use ADIC's DASADMIN software or manual means to clean the TLM drives, and use the tpclean command only for updating the drive's clean time. (All UNIX Media Manager Servers) Description: Could not add a disk storage unit with a UNC path via the bpstuadd command line. Workaround: A disk storage unit with a UNC path can be added with the NT Storage Unit Administration graphical user interface. (All NetBackup Servers) Description: The image files (.f files) in the NetBackup database are compressed prior to the time indicated in the "Delay to compress database" entry in the NetBackup configuration when using the "bpimage -compress" command. (All NetBackup Servers) Description: Provide fixes for Netbackup client support of Lotus Notes. (Windows NT Intel and Solaris2.6 NetBackup Clients) Description: Provide Tape Library Half-inch (TLH) robotic control on SGI IRIX. This includes support for the IBM 3494 robotic library using IBM's Automated Tape Library (ATL) interface. Additional Notes: This patch applies to IRIX 6.4 and higher, and specifically should not be installed on servers running IRIX 6.2. The robotic control uses a Library Name, as specified in /etc/ibmatl.conf, according to IBM ATL robotic control software. This IBM software must be installed on the IRIX server before the TLH robotic control is configured. See the Media Manager System Administrator's Guide - UNIX in Appendix G, "IBM Automated Tape Library (ATL)" for specifics on configuring and using TLH robotic control, following the Solaris model. (SGI Media Manager Servers) Description: Backups of very large numbers of files fail due to a 2 GB size limit on the files-file (the .f file). Additional Notes: This is for NetBackup Solaris 2.6 servers only. THIS WILL CAUSE NETBACKUP to fail if applied to a Solaris 2.5 server. (Solaris2.6 NetBackup Servers) Description: The backup images of HFS files systems from Mac OS X Server clients cannot be verified by the bpverify command. Shared memory alternate client restores of HFS files systems from Mac OS X Server clients cannot be done on media servers. These problems occur because the NetBackup "tar" command on media servers does not correctly recognize the HFS headers in some situations. (All UNIX NetBackup Servers) Description: If the bandwidth throttling feature is enabled for a range of IP addresses using the LIMIT_BANDWIDTH keyword in the bp.conf file or using the Configure NetBackup GUI, for some platforms it may enable bandwidth throttling for the wrong clients. Look in the bpsched debug log for the start_bpbrm line and check for the -throttle argument to bpbrm. If the -throttle argument is present, then throttling has been enabled for that job. If throttling appears to be enabled for the wrong clients, verify the clients specified in the LIMIT_BANDWIDTH entries are correct. If the clients are correct, this particular problem may be the cause. Note that this is only a problem for a range of IP addresses. Enabling bandwidth throttling for a single IP address with a LIMIT_BANDWIDTH statement should work correctly and is not effected by this change. Workaround: Enabling throttling on one IP address at a time instead of on a range of IP addresses. (All NetBackup Servers) Description: Improve loading time for tl8 and ts8 tape drives. (All Media Manager Servers) Description: NDMP sites with a large number of files may have insufficient resources or time to convert to the NetBackup catalog format. This patch allows the NDMP environment variable HIST to be set in the file list so that it may be set to "N". NetApp will then not generate the file history entries and the catalog will contain only the backup point. This eliminates the resources and time that would have been required to convert the file history. The syntax for the file list is SET HIST=N Additional Notes: Note that restore through NetBackup is then limited to the entire backup directory. Outside of NetBackup, manual tape positioning and a manual restore could done directly on the NetApp to restore a subset. (NetBackup servers hp700, hp800, rs6000, solaris, Windows NT Alpha and Intel) Description: This updates the /usr/openv/netbackup/dbext/install_dbext script, used to install the NetBackup Database Extension products. The name for the Post 3.2 release product NetBackup for Oracle Advanced BLI Extension was changed to NetBackup for Oracle on UNIX Advanced BLI Agent. Only NetBackup for Oracle on UNIX Advanced BLI Agent clients installed using the SUN pkgadd formatted NetBackup Database Extension products need this fix. The VERITAS ./install formatted NetBackup Database Extension products have this script as part of the package. (Solaris NetBackup servers) Description: The error "media read error" is returned for a phase 1 import operation or a Media Contents report when an image spans media on the AIX platform. The following error is seen in the import log: 09:40:04 INF - Found existing image information for client peony, class test_class, and schedule full_bkups, done on 01/04/00 16:28:20. 09:40:05 INF - Recreating copy 1, fragment 3 information for image. 09:40:10 INF - Status = media read error. The following error is seen in the Problems report: 01/05/00 09:18:21 peony - read error on media id A00001, drive index 2 reading header block, There is an input or output error. The following is seen in the bptm log: 09:40:09 [19116] <8> io_read_block: read error on media id A00001, drive index 2 reading header block, There is an input or output error. 09:40:09 [19116] <2> log_media_error: successfully wrote to error file - 01/05/00 09:40:09 A00001 2 READ_ERROR Workaround: The "media read error" may be ignored on an AIX platform when performing a phase 1 import or Media Contents report on media with an image that spans media. These operations successfully completed. (NetBackup rs6000 servers) Description: This change adds support for Fujitsu M8100 stackers. They are only supported on Solaris 2.6 and Solaris 7. The M8100 stacker must be set up to run in "System Mode" and "2LUN Mode". The following entries must be added to the st.conf file: "FUJITSU M8100", "Fujitsu M8100 1/2 Inch Cartridge", "FJ-M8100", FJ-M8100 = 1,0x24,0,0x1d63d,4,0x0,0x0,0x0,0x0,3; This change adds support for the ATL P3000 PassThru robot. Testing was done with firmware revision 1.41. The P3000 PassThru is supported as a TLD robot. There are 3 know limitations with Media Manager displays: 1) GUI and command line output for slot information does not line up in columns properly when slot numbers are greater than 999. There is no workaround for this display problem. 2) The NT Media and Device Management GUI does not sort volumes with slot numbers greater than 999 properly when the Slot header is clicked on. Use the Motif or Java GUIs to display volumes sorted by slot numbers. 3) There is a limit to the amount of information that can be displayed in the Robot Inventory Results window of the NT Media and Device Management GUI. When the limit is reached, no further output will be sent to the Results window. The selected operation will continue to completion. While the operation is in progress, the Start button will be gray. Using the Clear button between inventory operations will avoid the problem for smaller robots, but the problem can not be avoided when doing a "Show Contents" or "Compare Contents" operation on a large robot. Use the Motif or Java GUIs to display Robot Inventory Results. (All Media Manager servers) Description: Feature: For Tru64 (DEC Alpha), provide application error recovery for backups/restores. This error recovery will attempt to recover from intermittent fibre channel conditions that have nothing to do with the media itself. (NetBackup Alpha servers) Description: Enhance the -altreadhost feature that was first introduced in patchJ0820260 to also function when doing MPX duplication. (All NetBackup servers) Description: NetBackup may incorrectly deassign tapes that still contain valid images if the one of the following conditions is true: 1. There is a system resource problem during the bpexpdate -deassignempty process. This process verifies that each valid media has at least one image fragment in the image database. If a system resource problem occurs while scanning the image database such that some images are not found, a tape could be incorrectly deassigned. In essence, this failure would be the result of a malloc() failure on the master server during the processing of this request. It is unlikely that the processing of this request would continue under these circumstances, but if it did continue it is possible that it could lead to incorrectly deassigned media. 2. There is more than one storage unit defined for a media server, and the specified hostname for each storage unit represents a different network interface. For example, the hostname for storage unit A is specified as "xyz-nic1", which represents one network interface, and the hostname for storage unit B is specified as "xyz-nic2", which represents a different network interface on the same media server. Also, on the bpexpdate command, the -ev option was allowed to be entered with the -recalculate option. This leads to unpredictable results and is no longer allowed. The All Log Entries report would show entries similar to the following: media id A00000 removed from media manager database (manual deassign) Deassigned media-id A00000 on host xyz-nic2 media id B00000 removed from media manager database (manual deassign) Deassigned media-id B00000 on host xyz-nic2 Deassigned 2 out of 2 media-ids found. However, the Images on Media report would show that images are still present for one or more of the expired media. Workaround: This problem can be avoided by: On UNIX systems, create the file /usr/openv/netbackup/bin/bpsched.d/CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. On an NT server, create the file \NetBackup\bin\bpsched.d\CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. This will cause NetBackup to behave as it did in the 3.1.1 release. (All NetBackup servers) Description: For the date Jan. 3rd, 2000, log file log.0103100 was created instead of log.010300 Workaround: 1) Don't create the debug log directories /usr/openv/netbackup/logs/bpdbjobs or logs/nbmon, logs/xnb, logs/xbpadm, or logs/xbpmon. 2) Live with a strange filename. This *may* require periodically erasing old log files by hand if the automatic cleanup procedure doesn't find these files. Additional Notes: These debug logs contain internal debug information, and is not meaningful to anyone else. The impact of not having this debug output is very minor. (All NetBackup servers and clients) Description: The query to bptm for count drives now honors the BPTM_QUERY_TIMEOUT flag. This was hardcoded to 60 seconds. The default for BPTM_QUERY_TIMEOUT is 8 minutes. Patch J0820326 added functionality to allow jobs to requeue for storage unit not available errors and wait for the storage unit to become available again. This is not functional on NT. The reread storage unit functionality has been changed to always be enabled. This is to detect when drives are in use for non-backup purposes. It now uses a message queue for communication instead of shared memory. Bpsched will now recognize that a drive is in use for non-backups purposes (such as restores, duplicates, ...) and not attempt to use the drive. This is to prevent attempts to mount a tape on a drive that is in use, which will eventually cause a media mount time out. bpsched reports 96 error when a 219 error should be reported when the disable count media functionality is set. If bptm reports no media available, count media functionality is now re-enabled until the bpsched session is complete. (All NetBackup servers) Description: On HP10.2 and HP11.0 systems, autofs file systems are backed up even though the "Follow NFS" class attribute is not enabled. Workaround: Use exclude lists on the autofs file systems. (NetBackup clients HP10.20 and HP11.00) Description: Provides a workaround for a Solaris tape driver (st) bug where an attempted access of a drive when no tape is loaded, leads to a condition where the drive remains SCSI reserved. When the drive remains SCSI reserved, attempts to access the drive from other hosts will fail, due to a system error such as "write protected or reserved". If a bridge is rebooted or some other error situation occurs while using the passthrough driver, the avrd process will resort to using the st driver in order to scan the drive. With the st bug, the drive may be left SCSI reserved. If an administrator resets a DOWN'd drive, ltid will attempt to access the tape drive, which may leave the drive SCSI reserved, again because of the st bug. Workaround: Disable SCSI reserve/release in the Solaris st drive configuration entry for all configured multihosted drives. See ST_NO_RESERVE_RELEASE in the st(7D) man page on Solaris. (Solaris Media Manager servers) Description: The size of the EXB_RES_DRIVE_DESC was increased to support library serialization with tl8 robotics. (All Media Manager servers) Description: Provide NetBackup client support for machines running FreeBSD on Intel platforms. The binaries were built and tested on FreeBSD kernel version 3.3. The Motif interface (xbp) is not supported for this client. Encryption is not supported at this release of the FreeBSD client. (NetBackup INTEL/FreeBSD client) Description: Provide NetBackup client support for machines running MACINTOSH, Mac OS X Server. This client resembles typical UNIX NetBackup clients rather than the Mac OS NetBackup client. (NetBackup MACINTOSH/MacOSXS client) Description: Provide a means to define NetBackup clients of hardware/os types: Solaris/Solaris8, Solaris/Solaris_x86_8 and Solaris/Solaris_x86_7. These client platforms are supported with the Solaris/Solaris2.6 and Solaris/Solaris_x86_2.6 NetBackup 3.2GA binaries. (NetBackup Solaris clients) ============= patchJ0820326 ============= Description: Tar is not closing file descriptors when there are some HSM related problems. This can cause tar to reach the limit on the number of open file descriptors. This may cause an error message like : 09:21:35 (75372.001) WARNING tar can not get HSMNAME for /pathname/filename.c (All UNIX NetBackup clients) Description: A command used by the Java GUI is returning host names truncated to 12 characters. This causes the failure of GUI operations which rely on a volume's robot host field for host comparisons. For example, manually adding a volume in Media and Device Management would fail with the following error: "Could not add media ID xxx: robot number and robot host mismatch (61)." (All Media Manager Servers) Description: In the case of restoring migrated files on Solaris (Dmapi or Pain), add a check to verify that HSM is installed on the system. If it is not installed terminate the restore and issue the following message in the progress log: "Attempting to restore migrated files on a system with no HSM installed." (Solaris2.6, Solaris2.6 DMAPI, Solaris2.5 and Solaris2,5 DMAPI NetBackup clients) Description: Between the time a directory is read and the files are processed, files that are deleted will get: ERR - Cannot lstat /pathname/filename.c Errno = 2: No such file or directory These errors will cause the NetBackup catalog DB backups to fail and will cause partial success to be reported for normal filesystem backups. (All UNIX NetBackup clients) Description: When the bpbackup command is initiated with -i and -w options, if the job was killed by the job monitor while in the queued state, status 0 was returned instead of status 150. (All NetBackup Servers) Description: When creating a new NT class with bpadm, the prompt to set disaster recovery looks like this: disaster recover? (y/n) (y): If carriage return is hit to accept the default value, the disaster recovery field is set to the value of n. This change corrects the displayed default value: disaster recover? (y/n) (n): Workaround: Enter y for the value of disaster recovery. (All UNIX NetBackup Servers) Description: Empty lists in the configuration such as the Include or Exclude files list might not be displayed correctly by nbcfg (NT only), nbwin (NT client) or bpgetconfig (used by Java server). Instead, garbage or all but the first character of another configuration entry could be displayed. (All NetBackup Servers) Description: Add support for the following vendor/product inquiry strings: "QUANTUM L500" "IBM 7337" "IBM 3447" Without this patch, a "QUANTUM L500" robot with any empty slots will not initialize properly. The system log will show these errors: tldcd[xxxxx]: TLD(x) issuing initialize_element_status, slot x asc=0x3a, ascq=0x80 tldcd[xxxxx]: TLD(x) cannot clear slot x error, slot asc=0x3a, ascq=0x80 tldd[xxxxx]: TLD(x) going to DOWN state, status: Unable to initialize robot Without this patch, an "IBM 7337" or "IBM 3447" robot will not be recognized as the proper type. The system log will show the following information: tldcd[xxxxx]: TLD(x) Protocol is SCSI, type 11 tldd[xxxxx]: DecodeQuery() Actual status: STATUS_SUCCESS tldd[xxxxx]: TLD(x) going to UP state In the line above, "type 11" is incorrect. It should be "type 6". (All Media Manager Servers) Description: When doing a block incremental backup that was successful, bpend_notify recycled the output files before sending mail causing the mail file to be empty. Also, the post_checkpoint_notify script would restart the database when the DB was initially down when it should have left it in the down state when doing a SHUTDOWN_CKPT_RESTART backup. Workaround: Edit the scripts manually to move the sending of mail before the output file recycling in bpend_notify.oracle_bli. Look for the comments in the post_checkpoint_notify.oracle_bli script that say: # Restart the DB or put it in "end" mode. # If the backup method is SHUTDOWN_BKUP_RESTART, we don't do anything. and change the if statement that follows to: if [ -f $SHUTDOWN_CKPT_RESTART_FILE -a -f $SHUTDOWN_FILE ] (All NetBackup Servers) Description: When using NDMP, a robot daemon (tldcd or tl8cd) may run out of file descriptors due to a leak when the function ndmp_scsi_open_func is unable to open the robot. This can happen with a misconfigured robot in a Network Appliance OnTap 5.3+ host or it could happen when the network connection to a NDMP host is temporarily interrupted (for example network problems or a crashing filer). The syslog will contain messages similar to tldcd[337]: TLD(8) ndmp_scsi_open_func() failed until all file descriptors are consumed and then it will log a message similar to tldcd[337]: TLD(4) can't create child info pipe: Too many open files Workaround: When the error is caused by a misconfigured robot device, correct the configuration. (All Windows Alpha &Intel, HP700, HP800, RS6000, Solaris Media Manager Servers) Description: New feature: Add support for Media Manager authorization in the media manager commands and Media and Device Management GUIs. (All Media Manager Servers) Description: On HP_UX 10.20 servers, sending the server status to an NT client during a non-mpx backups can fail with a SIGPIPE. Workaround: Use multiplexed backups or turn on throttling for the NT clients, but use a large throttle Kbytes so that it doesn't have any affect. Additional Notes: If NetBackup is used with a firewall, applying this fix means ports 1025-5000 are used for backups/restores, an effect similar to enabling the multiplexing feature. (All NetBackup Servers) Description: Added support for the HP A5597A and HP A5617A. (All Media Manager Servers) Description: On Tru64 UNIX, in a fibre channel environment the logical unit number (LUN) must be passed along when submitting scsi tape requests to the Tru64 CAM driver. The behavior resulting from this omission is failed unmount requests and DOWN'd drives. On Tru64 UNIX 4.0f it was observed that it took a long time (5+ minutes) for robotic tape drives to move from AVR control to robotic control (i.e. TLD). Also, it took a similar amount of time to mount a tape. These two behaviors were caused by avrd reporting the drive ready incorrectly. This completes code modifications to support NetBackup and Media Manager on Compaq Tru64 UNIX 4.0f. (Alpha Media Manager Servers only) Description: If an attempt is made (by the operator or by the application) to DOWN a multihosted drive while a volume is being dismounted, the attempt will fail with "Drive is currently assigned". This fix queues the request so that it will occur after the dismount completes, when the drive is no longer assigned. Workaround: When downing a drive as the operator, wait until the drive is not in use before initiating the DOWN request. Additional Notes: This problem was introduced by changes made in patchJ0820260. (All Media Manager Servers) Description: 3.2GA NetBackup Authentication feature creates new directories on the NetBackup client. For new installations of NetBackup and for new NetBackup clients, these directories are created as part of the initial install. Upgrading NetBackup clients from 3.1.1 to 3.2, using the update_clients script. The update_clients script was failing because the directories, /usr/openv/bin and /usr/openv/var/auth, did not exist on the clients and the binaries that distribute the NetBackup client software did not create them. The fix provided in 3.2 patchP0820173 didn't work for 2 NetBackup client platforms, SCO/OpenServer and ALPHA/OSF1. (All UNIX NetBackup Servers and clients) Description: bpexpdate returns an error when trying to change the expiration date of Oracle images for a media id. (All NetBackup Servers) Description: When vsnDB is upgraded from 3.0(or earlier) to 3.2, the description field for each volume in an ACS, TLH, or TLM robot is incorrectly deleted. This will cause queries to vmd for volume information to fail. For example, a "vmquery -a -b" command will result in the error message "unexpected data received". Workaround: This problem only happens when upgrading vsnDB from 3.0(or earlier) directly to 3.2. A workaround is to upgrade to 3.1.1 and then upgrade again to 3.2. (All Media Manager Servers) Description: If a NetBackup job file contains NULL characters, jobs can not be deleted from the activity monitor. This is thought to occur when a filesystem is shutdown unexpectedly with unwritten buffers. Checks were added to handle jobs files with NULL characters, allowing jobs to be viewed with the activity monitor as in the DONE state and then deleted. Workaround: The job files can be manually deleted by changing to the netbackup/db/jobs directory and removing all files with that jobid. e.g., "rm -f 425 425.*" (All NetBackup Servers) Description: When Netware 4.11 buffers fill up, all network traffic is halted. bpdbjobs decreased buffer usage on all platforms to prevent the default Netware 4.11 buffers from filling up. Workaround: The Netware 4.11 Backup Exec server can be reconfigured with more buffers. (All NetBackup Servers) Description: When using TIR and streaming, a restore may result in a core dump. (All NetBackup Servers) Description: Teradata backup fails with server status 50 if the NetBackup server is a UNIX platform. Client connect failures for UNIX servers and Teradata clients when DbcExecuteCmd or DbcWriteFile routines are being used.Hold and release SIGUSR1, SIGUSR2, SIGCLD, and SIGTSTP around client connection (bpcd) calls. (All NetBackup Servers) Description: 1) If the client name is removed from a class after a worklist is built for a backup, but before the job gets queued to the main scheduler process, the main scheduler process may abort. This causes all active jobs to fail, ending in a 50 status. 2) If the allow multiple data streams option is selected in a class, the file list processing has problems with percent signs '%' embedded in the path name. This usually shows up as a 71 error, because the path name expands incorrectly and the files requested cannot be found. (All NetBackup Servers) Description: 1) A timing problem existed when starting MPX backups where the start of a new backup may be missed by bptm/bpdm. This could result in the backup timing out if other backups with large amounts of data were already in progress. 2) When doing MPX duplication, some progress messages still said "restore". New messages were added to say "duplication" when the operation was indeed a MPX duplication. (All NetBackup Servers) Description: Backup of a non-existent NDMP path causes bptm to run forever. Verbose bptm log will show no mover record progress over a long period of time. Workaround: Ensure that NDMP backup paths exist on the NDMP server. (Windows NT Alpha & Intel, HP700, HP800, RS6000, Solaris NetBackup Servers) Description: The robtest utility now recognizes up to 99 drives in a single robot. The fix is needed to support some of the larger pass-through robots such as the ATL P3000 Pass-Through. (All NetBackup Servers) Description: This fix corrects a possible problem in signal handling in the bpbrm process. Although a problem in this particular area has not been reported, an examination of the code exposed the possible problem. It is believed that if the problem was encountered, a backup job would fail during startup and then be retried successfully, effectively masking the problem. (All UNIX NetBackup Servers) Description: 1) When using SSO option, scheduler would not always use all drives available. 2) This fix is associated with the NCR Teradata extension. Since jobs of class NCR_TERADATA, named 'TD-NATIVE', are used internally by this extension and are not associated with any real backup or restore operation, this change removes these jobs from the activity monitor display. (All NetBackup Servers) Description: When process bptm is mounting a tape for bprecover, it was requesting write access. This has been changed to request read access only, so a write protected NetBackup catalog backup media can be successfully mounted. (All NetBackup Servers) Description: Provide Netbackup server and client support for hot backup/restore of Lotus Notes databases. Supported Netbackup clients include Solaris 2.6 and NT 4.0. (All NetBackup Servers, Solaris2.6 and Windows NT Intel NetBackup clients) Description: A sparse file that is just one large hole with a zero byte at the end has no real data to write during the restore, so the ftruncate() at the end fails with an I/O error. The progress log has the following error message: Couldn't truncate file /abc to 10000 bytes: I/O error (All UNIX NetBackup clients) Description: Added support for the DEC TL881 robot. (All NetBackup Servers) Description: Increase the number of LUN scanned to support a fibre channel/SSO environment. (Alpha NetBackup Servers only) ============= patchJ0820260 ============= Description: The NetBackup for Oracle encryption feature is restricted in the current release. This fix removes all known restrictions to NetBackup for Oracle encryption. (NetBackup clients that support NetBackup for Oracle:alpha, HP, Windows NT (Intel), ncr, rs6000, sequent, SGI, Solaris) Description: Encryption installation is unable to distribute or configure Linux/RedHat, SCO/UNIXWare or SCO/OpenServer clients for encryption. The command, /usr/openv/netbackup/bin/bpinst -CRYPT was missing table definitions for these 3 encryption client platforms, Linux/RedHat, SCO/UNIXWare and SCO/OpenServer. (All NetBackup servers) Description: When a client/user requests a login, and a creation of an instance of the user server which runs under the permissions of that user, ALL groups that the user belongs to will now be initialized for the user server instance. Prior to this change, only one (the login one) group would have been enabled. (All UNIX NetBackup clients) Description: The bpdbjobs -report command changed from 3.1 to 3.2 and the 3.2 bpdbjobs -report command did not recognize fields added to a report with the xbpmon command (generating a .xbpmonrc file). These added fields are now recognized and the correct report generated. (All NetBackup servers) Description: The bpjava-msvc server used the hostname of the client to keep track of the client, however, the hostname is not always available, and thus a failure to acquire it should not be an error. This change now uses the DNS address as a unique identifier for the client. Additionally, any error arising from the failure of a getpeername will now be reflected in the actual system error returned, v.s. the catch-all 'out of ports' message returned previously. Workaround: Make sure ALL clients/hosts have a name corresponding to their DNS address. (All UNIX NetBackup clients) Description: Before avrd attempts to open an NDMP tape drive it pings the NDMP host first to avoid downing drives inappropriately. If the ping is not successful the following message appears in syslog: ping of NDMP host (device ) timed out, skipping drive check While no additional log entries will be made until the NDMP host becomes available again, each cycle of avrd pings the NDMP host. When each ping fails the bug leaves a file descriptor open eventually consuming all available file descriptors. Workaround: A workaround is to correct whatever problem is preventing communication with the NDMP server and restarting the Media Manager demons. (Media Manager servers Windows NT (Alpha & Intel), hp700. hp800, rs6000, solaris) Feature Description: Provide a method where the read side of bpduplicate can be performed on a media server other than the one it was written on. This capability is available via bpduplicate commandline only. The new option for bpduplicate is -altreadhost . Note the media id must be available to the altreadhost for mounting. (All NetBackup servers) Description: For AIX, ignore ENOTREADY errno if it is returned from a MTOFFL ioctl(). The ENOTREADY can be caused by some STK drives returning a check condition on a SCSI unload, indicating the drive needs cleaning. (NetBackup rs6000 servers only) Description: Correct problem where tl8d could incorrectly believe there are no NDMP drives configured when there really is. (Media Manager servers Windows NT (Alpha & Intel), hp700, hp800, rs6000 and solaris) Description: When removing directories in /usr/openv/netbackup/tir_info, unlink is used instead of rmdir. This causes the inode to get lost until fsck is done on the filesystem. Workaround: If TIR move detection is on, don't turn it off. If TIR move detection is turned off, do the following as root on the client: rm -rf /usr/openv/netbackup/tir_info (All UNIX NetBackup clients) Description: When doing a raw partition backup an exit status of 1 will be returned if any data is written to the tape when errors have been encountered. (All UNIX NetBackup clients) Description: Tapes in slot 0 of Exabyte 210 and Exabyte 220 robotic libraries cannot be mounted. System/Event log entries will show the following, if VERBOSE is enabled in the Media Manager configuration (vm.conf) file: Sep 1 14:05:35 host1 tl8cd[8773]: TL8(3) robot is EXB2X0 compatible with barcode reader . . . Sep 1 14:06:15 host1 tl8cd[8773]: Processing MOUNT, TL8(3) drive 2, slot 0, barcode 00BCK047, vsn BCK047 Sep 1 14:06:16 host1 tl8cd[8798]: TL8(3) 0 is an invalid slot Sep 1 14:06:16 host1 tl8d[8766]: DecodeMount(): TL8(3) drive 2, Actual status: Requested slot does not exist in robot Workaround: Change the volume pool for any volume in slot 0 to be "None", provided the media has not been assigned, so that it doesn't get requested by the application. If a volume in slot 0 is already assigned, move it to another slot, move an unassigned volume into slot 0, change its volume pool to "None", and update the volume configuration accordingly. Additional Notes: If AUTO_UPDATE_ROBOT is enabled, media automatically injected through the Exabyte 210/220 mailslot may be placed into slot 0, if slot 0 is empty. The best workaround is to place an unassigned tape that does not need to be mounted into slot 0, and change its volume pool to be "None". (All Media Manager servers) Description: In NetBackup 3.1.x, UNIX servers and clients selected non-reserved ports using a decrement method from port number 5000. This allowed firewall administrators to only open a few port numbers below 5000. NT servers and clients selected non-reserved ports randomly between 1025 and 5000. In NetBackup 3.2GA, both UNIX and NT servers and clients used the random method. This required firewall administrators to open ports 1025 to 5000. That many open ports is generally unacceptable in a firewall environment. This patch causes UNIX servers and clients to use the 3.1.x decrement method by default. NT servers and clients will continue to use the random method by default. There is a new configuration entry available to change the default: RANDOM_PORTS = [YES | NO] The default for UNIX is "NO" and the default for NT is "YES". Workaround: Open ports 1025 to 5000 over the firewall. This generally is not acceptable. Additional Notes: There is a bug in NT networking that prevents NetBackup from using the decrement method. That's why NT continues to use the random method. NT clients will generally work over the firewall even with a limited number of open ports (e.g. 4900 to 5000), but NT servers will not. If a NT server is used over a firewall, ports 1025 to 5000 must be opened. (All UNIX NetBackup servers and all UNIX NetBackup clients) Description: The NetBackup Java Interface Backup tab does not correctly report the size of files larger than 2GB. (All UNIX NetBackup clients) Description: While running the NetBackup Client interface and attempting to submit a backup/restore request to a NetBackup server, the interface crashes. This problem is only seen when machines with multiple network cards are being used. Workaround: If authentication is not in use, remove the methods.txt file found in the var/auth directory. (All NetBackup servers) Description: Quick I/O symbolic links restored with RMAN proxy copy need special handling when renames are involved. (All UNIX NetBackup clients) Description: The HP-UX 11.00 bpbkar needs to be linked with the VxFS 3.3.2 libraries. (NetBackup client HP-UX11.00) Description: If the TS8 robotic daemon is used to control an Exabyte 210 with a barcode reader, backups will fail with the messages "Drive does not exist in robot." Workaround: Control the Exabyte 210 using the TL8 robotic daemon. (All Media Manager servers) Description: PatchJ0820226 provided several new entries for the bp.conf file and new functionality that allowed jobs to requeue on a required storage unit that was not available (status 219) This leads to two situations: 1) Jobs that were requeued on a required storage unit would be retried multiple times, even though the job had completed successfully. To kill one of these jobs, all the jobs in the job monitor have to be killed. 2) Jobs started using bpbackup with the -i and -w options would return status 0, even if the job failed. Also introduced by patchJ0820226, if a shared memory identifier of 0 was returned by the OS, a segmentation violation would occur. Additional Notes: Multiple jobs can be started with a bpbackup -i -w. This is because multiple clients can be specified on the class, or multiple streams enabled. The bpbackup command can get only one status. The change made will return the last non-zero status it finds on completion. If all jobs complete with a 0 status, then 0 status is returned. (All NetBackup servers) Description: In NetBackup 3.2, the NetBackup Java GUI application and its application server uses port 13722 for its initial communication - the login. Once logged in, a random free port was used for all subsequent communication. This made it virtually impossible to use the NetBackup Java GUI applications for remote administration in a firewall environment. This feature forces the subsequent port selection by the NetBackup Java application server to be the first one available decrementing from port number 5000 (the initial communication remains the same - use of port 13722). The number of consecutive free ports in addition to 13722 to 'open up' is dependent on usage and a sites security policies. Each execution instance of the jbpSA or jnbSA (i.e., the NetBackup Java GUI application startup) commands will require one of these ports. This feature allows one to configure a set of ports to allow administration from a machine outside the firewall, e.g., a Windows machine at home on which the NetBackup Java Display Console for Microsoft Windows software is installed. If the machine inside the firewall is a GDM, one has the ability to administer multiple NetBackup servers. The NetBackup Java application server will use the sequential port selection method as described above by default. However, if "RANDOM_PORTS" is specified in the bp.conf file, ports will be selected randomly in the range 1025 to 5000. Additional Notes: Additionally, this change implements a server timeout, which will prevent user (and transient master) servers from becoming zombies if, after they are created, no initial request is made within 60 seconds. (All UNIX NetBackup clients) Description: When trying to perform a restore, the restore may fail with a 'handshake failure' error which is caused by tar (the restore process) sending a message to bpbrm (the backup/restore manager) indicating that it received a parameter that it does not understand. (All NetBackup servers) Description: The UNIX tar complains about an unknown parameter (-J job_total=X). This message is sent to the server and may cause bpbrm to fail the restore. (All UNIX NetBackup clients) Description: Support the IBM 3590 E11 stacker as a TSH robot. Additional Notes: The IBM 3590 E11 stacker must be in Random 2-LUN ACF mode. For servers running AIX: On AIX, the IBM 3590 E11 robotics must be controlled by the VERITAS pass-through driver, ovpass. Please consult the Media Manager Device Configuration Guide - UNIX for instructions on how to install ovpass, as well as on how to configure SCSI robotic controls using the ovpass driver. The resulting robotic control path should be in the form "/dev/ovpassX" where X is the instance of the ovpass driver. vmconf or another device configuration interface, can be used to configure the device. NOTE: Do not use instructions in the section "Configuring IBM 3590 Stacker Robotic Controls" when configuring the IBM 3590 E11 stacker because that section only applies to the IBM 3590 B11 stacker, which has a "Random" ACF mode that is not supported by IBM in the E11 model. Instead, consult examples 1-4 in the documentation for details on creating the pass-through device file for the IBM 3590 E11 robotics. For servers running Solaris or IRIX: Use existing documentation or vmconf to configure the IBM 3590 E11 stacker as a TSH robot. (All Media Manager servers) Description Features: - For all platforms, improve the NetBackup tape selection algorithm such that there are not so many volume database queries required. This change will be most noticeable to installations with many non-FULL active tapes in the media database. - For NT and Solaris only, provide application error recovery for backups/restores. This error recovery will attempt to recover from intermittent fibre channel conditions that have nothing to do with the media itself. Problems: - For all platforms, fix problem where the importing of TIR images did not set the expiration date of the image correctly. - For all platforms, fix problem where bptm issues a misleading "unable to write bpbrm message" to the NetBackup error log after MPX restores. The error is misleading because it give the impression there was a problem with the restore, but the restore worked correctly. (All NetBackup servers) Description: If there are images in the NetBackup catalog that were created from a version of NetBackup prior to 3.2 that have fewer than 10 file entries in the image and if those images were compressed using the 'Compress NetBackup Catalog' option prior to 3.2, then those images will not be properly decompressed unless this fix is applied. While browsing the affected image for a restore, the following error may occur: WARNING: unable to obtain list of files using specified search criteria. Or the following message in the output from a verify operation: INF - Expected filename /dir/file1 in database, found no more files. Or the following in the bpdbm debug log file, if available: <2> xpress_image: only has 8 files - not compressing <16> list_files: cannot decompress (All UNIX NetBackup servers) Description: The script, /usr/openv/netbackup/dbext/install_dbext, has been updated for the post 3.2 release of several NetBackup Database extensions. This only affects the pkgadd installation of the NetBackup Database extensions. This version of install_dbext is required to install the Post 3.2GA releases of NetBackup Database extension products. (Solaris NetBackup servers) ============= patchJ0820226 ============= Description: When upgrading to 3.2 on AIX, the vsnDB and robotic_def files in volmgr/database are not properly converted to the new 3.2 format. The system log will show the following error when vmd attempts to convert vsnDB: vmd[4453092]: database auto-upgrade failed: database read operation read too few bytes The aix_mmdb_upgrade script will stop all the NetBackup and Media Manager daemons, move the robotic_def and vsnDB files back to their original locations before the failed upgrade (if necessary), convert the files, and restart the daemons. (RS6000 NetBackup/Media Manager servers only) Description: Restores that need a large incremental take a long time to figure out what files are needed from each image. Also, when a restore is killed, the display of the "was not restored" messages takes too long. Only 500 of them will now be displayed and a count of the number of additional messages that were not displayed. To see all the files that were not restored, create /usr/openv/netbackup/DISPLAY_ALL_NOT_RESTORED on the master server. (All UNIX NetBackup servers and Windows NT Intel and Alpha) Description: Provide support for robot serialization with the TLD robotic type. (All UNIX Media Manager servers and Windows NT Intel and Alpha) Description: bpretlevel did not operate correctly with the -M option. bpretlevel did not output an error message when the command failed. bpretlevel would get a segmentation violation if displaying a retention level that was set to expire immediately. bpretlevel did not work unless the -M option was used. (All UNIX NetBackup servers and Windows NT Intel and Alpha) Description: There were a number of issues having to do with network problems and slave servers not responding. The symptoms would be a 219 error and scheduler processes that appear to be hung. Changes were made to bpsched to handle timeouts to the slaves better. Also additional functionality was added to allow jobs queued to a storage unit to wait for a 219 error to clear instead of failing. This way jobs don't loose a try and also don't have to wait to be rescheduled. There are new configuration parameters added to bp.conf to support this. WAIT_IN_QUEUE If this flag is present in the bp.conf file, a job that is queued and finds that it's required storage unit is unavailable, the job will go into the requeued state with a 219 status and wait for the storage unit to become available again. QUEUE_ON_ERROR If this flag is present in the bp.conf file, a job that is ready to be queued but has does not have a required storage unit available, will be placed on the work queue in the requeued state with a 219 error. TIMEOUT_IN_QUEUE This value set in the bp.conf file sets up how long a job is allowed to wait on a queue in the requeued state for a 219 error to clear. The default value is 10 hours. RE_READ_INTERVAL This value in bp.conf is used when jobs are waiting in the workqueue for a 219 error to clear up. bpsched will periodically poll the unavailable storage unit to see if it has become available again. The default value for the RE_READ_INTERVAL is 5 minutes. BPTM_QUERY_TIMEOUT This value is a timeout value used when bptm is doing a count media or a count of the number of available drives. The default value is 8 minutes. (All UNIX NetBackup servers and Windows NT Intel and Alpha) Description: When doing a TIR restore, the display of "is not in the true image list. Skipping." messages can take a long time. Limit the display to 1000 messages or create the file, /usr/openv/netbackup/DISPLAY_ALL_NOT_RESTORED on the master server. (All UNIX NetBackup servers and Windows NT Intel and Alpha) Description: Enhanced configuration analysis for the case when a single drive is specified on the command line. This is the way the MHD wizards call this command. (All UNIX Media Manager servers and Windows NT Intel and Alpha) Description: The usage message for /usr/openv/volmgr/bin/sg.build was not correct. (Solaris Media Manager servers only) Description: The listen backlog is set to 5 on NCR which causes connection timeouts when more than 5 are pending. (NCR NetBackup servers only) Description: A directory on a local file system is incorrectly considered an NFS file system during backup and is not backed up if 'Follow NFS' is set to 'No'. The following type of warning message can be seen in the logs. WRN - /export/data10/qqq is on an NFS file system. Skipping. This situation can occur on IRIX when automount is used with a direct map. This occurs when a machine is exporting a file system and a directory on that file system is in the auto.direct file. For an example, the problem occurs on the machine cobra when the following is true: /dev/dsk/dks1d7s7 is an xfs file system mounted on /export/data10 AND auto.direct contains the following: /data10/qqq -rw,intr cobra:/export/data10/qqq (SGI/IRIX6 NetBackup clients) Description: Netbackup binaries fail with "can't find libnfs.so" on Sequent systems in which the libnfs package is not installed. On Sequent, the NFS option is a separately licensable add on package. (Sequent4.2 NetBackup servers and clients) Description: vmd shuts down due to too many file descriptors open, after accumulating numerous failures trying to obtain a connecting host name. The following message may appear in the system log: vmd[2006]: accept() failed, Too many open files The volmgr/daemon debug log may have the following message: <3> peer_hostname[25494]: gethostbyaddr() h_errno: (1) <3> listen_loop[25494]: unable to get peername, errno=4 Workaround: 1. fix the network 2. stop/restart vmd periodically as preventative maintenance 3. tune the operating system so that a process can have many more file descriptors opened (All UNIX Media Manager servers and Windows NT Intel and Alpha) Description: Add device support for the Compaq ESL9000 robotic tape library. (All UNIX Media Manager servers and Windows NT Intel and Alpha) Description: If a tape is stuck in a drive due to a hardware or media error, the media cannot be dismounted (as expected), but the TLM robotic daemon continues trying, instead of configuring the drive DOWN. The following system log entries may occur: tlmd[6958]: TLM(2) DismountTape 50128D from drive 2 tlmd[10338]: TLM(2) open failed in io_open, I/O error tlmd[10338]: TLM(2) open failed in io_open, I/O error tlmd[10338]: TLM(2) open failed in io_open, I/O error tlmd[10338]: TLM(2) dismount failure for volser 50128D on drive 2 (DE04), d_errno = 12, The drive did not eject a volume. tlmd[6958]: DecodeDismount(): TLM(2) drive 2, Actual status: System error occurred during robot operation tlmd[6958]: Unexpected response status (30) in DecodeDismount (All UNIX Media Manager servers) Description: The release of NetBackup for Teradata Database Extension requires changes to NetBackup binaries. NetBackup for Teradata Database Extension is patchF0820222. (All UNIX NetBackup servers and Windows NT Intel and Alpha) (NCR and Windows NT Intel NetBackup clients) Description: Provide fixes and enhancements for Multihosted Drives (Shared Storage Option): - Correct problem where standalone multihosted drives were not always getting allocated properly. - Enhance error recovery, accounting for situations where machines crash or are rebooted with drives reserved. - Add multihosted drive support to robotic libraries using TL8 robotic control. - Modify Multihosted Drives design so that only one host requires I/O access to a tape drive at a specific point in time. This allows vendors to provide customized scripts to support dynamic hardware configurations. NOTE: The multihost_drive_notify script has been modified to support this design change. - Place an upper limit of 180 seconds on AVRD_SCAN_DELAY, since higher values may adversely affect the functionality of Media Manager. (All UNIX Media Manager servers and Windows NT Intel and Alpha) **Description:** THIS ADDRESSES A PROBLEM WHERE DATA LOSS CAN OCCUR. Every 12 hours, the bpexpdate -deassignempty command is started, which compares the media database to the image database and deassigns media that no longer contain valid images. If backups are in progress during this operation, NetBackup may incorrectly deassign a tape that contains a valid image if one or more fragments of the backups span media. This occurs in the situation where the spanned media only contains fragments for backups that are still in progress. If a tape is deassigned incorrectly and then reused, a restore of the affected image will result in a 94 error (cannot position to correct image). The following entries are seen in the "all log entries" report prior to the restore attempt: bpexpdate Deassigned media-id xxxxxx on host srvr1 bpexpdate Deassigned 1 out of 99 media-ids found. where the media-id xxxxxx contains one of the fragments of the image to restore. Another symptom of this problem is that the Images on Media report does not match the Media Contents report for the tape that is incorrectly deassigned. Workaround: This problem can be avoided by: On UNIX systems, create the file /usr/openv/netbackup/bin/bpsched.d/CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. On an NT server, create the file \NetBackup\bin\bpsched.d\CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. This will cause NetBackup to behave as it did in the 3.1.1 release. After this patch is applied, remove the CHECK_EXPIRED_MEDIA_INTERVAL file. (All UNIX NetBackup servers and Windows NT Intel and Alpha) ============= patchF0820201 ============= Description: This is the feature release of NetBackup Intelligent Disaster Recovery (IDR) for NetBackup Intel Windows NT clients. IDR is supported on all NetBackup server platforms with Intel Windows NT clients. This patch is the server part for all supported UNIX NetBackup servers. In order to fully protect a network with Intelligent Disaster Recovery (IDR), a class needs to be configured to collect disaster recovery (DR) information from the clients. The option to collect DR information is only available for classes of type MS-Windows-NT. After this option is enabled, DR information will be gathered for every client in the class after every backup. Currently, this option can only be enabled via 'bpadm' or via the Windows NT version of the Backup Policy Manager. The option is currently not available in 'xbpadm' or the java version of the Backup Policy Manager. Any NetBackup Windows NT client protected by IDR must have its C: drive shared and the first partition of the first drive must be the C: drive. A NetBackup Intel Windows NT server must be a client to another NetBackup server's IDR class to be protected with IDR. Install patchJ0820226.winnt.intel.exe on the Intel Windows NT NetBackup servers and clients. Install patchJ0820226.winnt.alpha.exe on Alpha Windows NT NetBackup servers. See patchJ0820226.README.NT for installation instructions. ============= patchJ0820198 ============= Description: Scheduled encrypted Oracle backups fail. This problem might affect some encrypted file system backups (all UNIX NetBackup clients) **Description:** THIS ADDRESSES A PROBLEM WHERE DATA LOSS CAN OCCUR. NetBackup may incorrectly deassign tapes that still contain valid images. This may occur if the following is true: - There is more than one storage unit defined for a media server, and the specified hostname for each storage unit is not identical. For example, the hostname for storage unit A is specified as "xyz" and the hostname for storage unit B is specified as "xyz.mydomain.com". The All Log Entries report would show entries similar to the following: media id A00000 removed from media manager database (manual deassign) Deassigned media-id A00000 on host xyz.mydomain.com media id B00000 removed from media manager database (manual deassign) Deassigned media-id B00000 on host xyz.mydomain.com Deassigned 2 out of 2 media-ids found. However, the Images on Media report would show that images are still present for one or more of the expired media. This patch also prevents media from being deassigned if a storage unit is defined for a remote media server that is incorrectly configured as a master server. Workaround: This problem can be avoided by: 1) Assuring that the storage unit definitions for a server which has more than one storage unit use the same name (not aliases for each other). OR 2) On UNIX systems, create the file /usr/openv/netbackup/bin/bpsched.d/CHECK_EXPIRED_MEDIA_INTERVAL On an NT server, create the file \NetBackup\bin\bpsched.d\CHECK_EXPIRED_MEDIA_INTERVAL and enter a 0 (zero) on the first and only line of the file. This will cause NetBackup to behave as it did in the 3.1.1 release. After this patch is applied, remove the CHECK_EXPIRED_MEDIA_INTERVAL file. (All NetBackup Server platforms) Description: NDMP 3-way backup intermittently saved inaccurate block counts in the catalog resulting in later verify failures. NDMP restore of very large directories could inappropriately timeout after 5 minutes. (alpha-nt, hp700, hp800, intel-nt, rs6000, solaris) Description: Executing xbp on Linux client will core dump with a segmentation fault if the list of files to backup or restore is greater than 1. Workaround: use the curses interface (bp). (Linux/RedHat NetBackup client) Description: - Add support for the following peripherals using TL8 robotic control: o Exabyte X200 o ADIC Scalar 1000 (AIT drives only) This includes automated media injection for TL8 robots. - Make drive scan delay configurable. Additional Notes: Media Manager now supports the following vm.conf options: PREVENT_MEDIA_REMOVAL AUTO_UPDATE_ROBOT DISABLE_USE_SEED AVRD_SCAN_DELAY = Behavior: By default, TL8 robots implement the use of SCSI allow/prevent as a security mechanism to prevent unauthorized access to the robotic library. The use of SCSI allow/prevent is now disabled by default, in order to facilitate automated media injection. Use PREVENT_MEDIA_REMOVAL to enable SCSI allow/prevent and restrict access to the library. Use AUTO_UPDATE_ROBOT to enable automated TL8 mailslot scanning, movement of media from mailslot to available slots in the library, and automated update of the volume database after mailslot use, or after the robot access door has been opened and closed. By default, automatic updates from AUTO_UPDATE_ROBOT will use a seed for defining media that do not have readable barcodes. If site policy dictates that all media must have readable barcodes, add the DISABLE_USE_SEED option. If specified, inventory update will fail when no barcode is present. Use AVRD_SCAN_DELAY to configure the number of seconds that the AVR daemon (avrd) will sleep time between drive scan cycles. The default is 15 seconds. (All Media Manager platforms) Description: If "bpimage -create_image_list -client " was used to speed up image operations, restores can lock the file for an extended period of time and cause backups for the client to fail when trying to create the image file. This fix releases the lock between image list searches. Workaround: Delete the /usr/openv/netbackup/db/images//IMAGE* files. (All NetBackup server platforms) Description: On a busy system, there can be temporary file name conflicts on a UNIX platform and authentication failure can happen between that UNIX platform and itself or another host. Additional Notes: This fix was included in 3.2GA for Windows 95/98/NT. (All UNIX NetBackup clients) Description: 1.) Solaris workstation configured with lun 1 devices enabled, may list certain lun1 robots listed twice, after vmconf is run. 2.) vmconf now scans the system for devices automatically on SGI platforms. 3.) The sgscan utility now recognizes the IBM 3590 as a changer as well as a drive on solaris. (Media Manager sgi, solaris platform) Description: Prevent SCSI timeouts on robotic/tape operations that take more than 60 seconds. This can occur only on systems running Solaris 7 in 64 bit mode. Additional Notes: Valid on Solaris 7 in 64 bit mode only. (NetBackup and Media Manager solaris platform) Description: When using bp to restore files, the class type cannot be changed to either "MS-SQL-Server" or "MS-Exchange-Server". If either of these are selected, the class type is changed to "Standard". When using bp with a keyword to browse for restores, bp fails to find file(s) with the keyword match. Workaround: The xbp or java interface can be used to avoid this problem. (All UNIX NetBackup clients) Description: Robot inventory fails for ts8 robots. Workaround: For Exabyte 220, robot can be configured as tl8. There is no workaround for Exabyte 210. (All Media Manager platforms) Description: A NetBackup server may hang during a restore for a FlashBackup client if indexing is enabled. When it does hang, the restore is logged as an Auspex backup in the bprd log. A NetBackup Solaris server with Auspex FastBackup clients will be unable to restore the Auspex FastBackup clients due to a missing library, libsfr.so ============= patchP0820173 ============= Description: Upgrade of 3.1.1 clients to 3.2 NetBackup client software fails. The update_clients script reports the server doesn't have Encryption enable or update_clients script fails and the /tmp/update_clients. log has the error message, Can't openv /usr/openv/bin/vautl_util on client XXX When asked during an upgrade installation of NetBackup: "Do you want to update the NetBackup software on the clients?" Answer no and install this patch. NetBackup 3.2GA includes new files and directories for the authentication feature that don't exist on a 3.1.1 client. The update_clients will fail since these directories don't exist on the client. bpinst now makes the /usr/openv/bin and /usr/openv/var/auth directories on the client. The update_clients script creates a /tmp/bp directory on the client. If there was a error, this directory was not removed and subsequent attempts to update the client failed because the /tmp/bp directory exists. The /tmp/bp directory is now removed on the client when update_clients fails. The Errata sheet with NetBackup 3.2GA describes a workaround for Client Software Installation Problem, adding an "ALLOW_ENCRYPTION" to the server's bp.conf. This workaround is no longer needed. The Errata sheet also describes a workaround for, Client Software Install Script Change for HP-UX and Solaris. Install this patch before new clients are installed to eliminate the need to edit the files described in the workaround. The script ftp_to_client and install_client were changed to correct the handling of the JAVAnbj.conf file in tmp (don't rm it too soon) and correct Solaris7 use. (All UNIX NetBackup servers) README -- Last modified date: Wednesday, March 13, 2002