Patch Name: PHSS_20083 Patch Description: s700_800 10.X OV OB3.00 patch - CS packet Creation Date: 99/10/08 Post Date: 99/10/20 Hardware Platforms - OS Releases: s700: 10.00 10.01 10.10 10.20 s800: 10.00 10.01 10.10 10.20 Products: OmniBackII A.03.00 Filesets: OMNIBACK-II.OMNI-CS Automatic Reboot?: No Status: General Superseded Critical: No Path Name: /hp-ux_patches/s700_800/10.X/PHSS_20083 Symptoms: PHSS_20083: Id = NSMex02773, HSLco17205 Option '-change_objname' is obsolete, but is still explained in the usage and recognized by the omnidbutil executable. If used, the omnidbutil fails with the message: "Invalid old object name specified!" "Update failed!" Two objects are listed in the restore GUI - one with capital and one in small letters. PHSS_19938: Id = NSMbb20735 1. Backup pre-exec script for host object should be executed once, before expanding the host into filesystem objects, and post-exec script should be executed once, after all the filesystem objects for that host have completed. Instead, the pre/post-exec script is started for each filesystem object expanded out of a host. 2. In case the following general release patches are installed on the Cell Manager system, PHSS_18697, PHSS_19359 (10.x) PHSS_18698, PHSS_19360 (11.x) the pre/post-exec script for a host backup is not executed at all. Id = NSMex02698 Backup of the OmniBack II internal database fails with an error: "Connection reset by peer." PHSS_19359: Id = HSLco15992 Backup of a file bigger than one tape segment (default 100 Mb) fails with Major error message: [61:1005] Got unexpected close from BMA on [61:1005] Got unexpected close from BMA on [61:1005] Got unexpected close from BMA on PHSS_18697: Id = NSMex02413 OmniBack II requires '61-250 Slot Libraries Extension' license if MO library with more than 30 slots is used. Id = NSMbb24210 Session pre-exec script fails, and then the post-exec script is not executed. Id = NSMex01842 Writeascii always starts with the first host in the cell when restarted, even if it has already completed for some hosts. This is very time-consuming operation. Id = NSMex02098 OmniBack II internal database backup reports a critical error message: "Velocis server ob2: (Change log cycle timeout)." Id = NSMex02330 Command 'omnisv.sh start' reports error, if the file mmd.ctx is not present, when trying to remove it. Id = NSMex02510 Host backup does not work correctly if name of one specified mountpoint is subset of another specified mountpoint. Id = NSMbb22383 MC/SG configuration doesn't add the package hostname into the cell_info file. Id = NSMex01965 "omnidbutil" return code is "0" even if it fails. Id = NSMbb22689 MMD daemon crashes with core-dump when GRAU with two or more drives is used. Id = NSMex01889 All sessions are marked as 'Full' after the writeascii/ readascii operation. Id = NSMex01998 Informix restore of Logical Log reports the error message: "Some of the backup devices are occupied. Session is waiting for all the devices to get free." Id = NSMex02146 MMD reports "Host for library DLT7000-Library not found." if hostname where the library is configured is longer than 32 characters. Id = NSMex02117 Licenses created for IP address that contains '0' are not accepted. Id = NSMex02009 Cell Administration GUI hangs when installing or uninstalling clients. Id = NSMex02242 Directories are backed up even if excluded in host backup. Id = NSMex02303 If a host backup object cannot be contacted in a multiple object backup specification, the backup exit status is 10 (Completed/Errors) instead of 11 (Completed/Failures). Id = NSMex02218 Backup hangs if one of the client systems is overloaded. Id = HSLco14785 Backup of an object fails with an error: "Got unexpected close from Unknown on ." Id = NSMex02351 If 'UpgradeIncrToFull' global parameter is set to 0, the backup of the client is still started. PHSS_17299: Id = NSMbb21666 If host backup uses the option '-tree ' for UNIX disk agent client, the specified directory is not backed up unless it is a mount point. Id = NSMex01657 If restore of large number of files is started via GUI, some files are not restored, although the session completed successfully. Id = HSLco11971 'omnidbutil -chkdbspace' doesn't report dbspace low, although space available is lower than the global option DbFreeExtFileSpace specifies. Id = NSMex01913 omniforsg.ksh configuration script failing for primary system. processes remaining on the system, all with the crs' pid as the parent process. Id = HSLco12167 omniforsg.ksh adding trash to /etc/opt/omni/cell/cell_info file for secondary system Id = NSMbb22116 failover doesn't restart backup sessions although the backup specification contains 'CLUS_RESTART FAILED' or 'CLUS_RESTART ALL' keywords. The package is halted. Id = NSMex01889 Omnidbutil -readascii fails with an error: "bad id in file : "pos.txt", line#: , field#: 2!" Id = HSLco12023 If several hosts are to be backed up and one or more of them cannot be connected when the backup is running, then the backup session exit status is 10 (Completed/errors). Id = NSMex01884 The crs daemon dying. Id = HSLco11955 Restore doesn't work after "List From Media". Id = NSMex01965 The return value of "omnidbutil" is always "0". Defect Description: PHSS_20083: Id = NSMex02773, HSLco17205 The option '-change_objname' for the command has been introduced with the SSPUX300_013. It modifies the object name in the OB2 internal database. This results in restore from the GUI not being possible for such object, but it can still be restored from the CLI command 'omnir'. From the Cell Manager patch PHSS_18697, the option has been considered obsolete, but is still explained in the usage, and recognized by the command 'omnidbutil', although it doesn't change the object anymore, and fails instead. OmniBack II A.03.00 resolves all hostnames to small letters at the backup time, but old objects, backed up with OmniBack II A.02.55 show up as capital letter hostnames. The result is one object being listed twice, although a detailed look into the objects shows that the versions are different. The restore of both objects is possible from the GUI, but from the command line interface only the small letters object can be restored. Resolution: The option '-change_objname' is removed from the usage and is not recognized by the 'omnidbutil' command anymore. PHSS_19938: Id = NSMbb20735 When a host object is expanded into filesystems at the backup time, the pre/post-exec script is copied into each filesystem startup parameters. This causes the pre/post-exec script to be started for each filesystem mounted to one host. This is a defective behaviour. With the patches PHSS_18697, PHSS_19359 (10.x), PHSS_18698, PHSS_19360 (11.x), the filesystem pre/post-exec scripts are initialized to NULL string for host object, and no pre/post-exec scripts are started, which is also a defective behaviour. Resolution: The bsm now correctly expands host object into filesystem objects, and starts one pre/post-exec script per host. Id = NSMex02698 Filenames of OmniBack II internal database files get corrupted in the db disk agent record, which causes the db disk agent to core dump. This results the backup session to fail with an error message: "Connection reset by peer." Part of SSPUX300_084 Resolution: Db backup disk agent now handles the filenames correctly. PHSS_19359: Id = HSLco15992 Improper handling of empty catalog message, when one file spans over 3 tape segments, causes backup session manager to abort instead of continuing with the backup. Resolution: Backup session manager now correctly responds to an empty catalog message from the media agent. PHSS_18697: Id = NSMex02413 If MO library is used, each side of a platter is considered as one slot. This requires a slot extension license, even if there are not more than 60 platters inside the library. Resolution: Slots are taken per platter now, not per side as before. Id = NSMbb24210 The pre/post-exec scripts on a session basis are usualy used for shutting down/starting up user's databases, which are then backed up as a filesystem or rawdisk. When a session pre-exec script fails, it is usualy required that the post-exec is executed anyway in order to get the user's database back alive. Resolution: Backup session manager now starts the post-exec script even if pre-exec has failed. Id = NSMex01842 Writeascii and readascii operations are always done on the complete set of hosts in a cell. This can take a lot of time, if internal database is big, and not all hosts need to be in writeascii/readascii process. Resolution: With this patch, a special functionality is implemented for selecting a list of hosts to perform the writeascii/readascii for. See the Special Installation Instrcutions for more details on how to select the list of hosts to be processed by writeascii/readascii. ID = NSMex02098 OmniBack II internal database backup disk agent gives up trying to backup the database, if it is not accessible. See Special Installation Instructions for more info on how to set the number of retries. Resolution: User configurable number of retries implemented. Id = NSMex02330 The script omnisv.sh doesnt't check for the presence of the file mmd.ctx before it tries to remove it. If the file is not there (no MA started yet), a system error will be reported. Resolution: Additional checking for the presence of the mmd.ctx file implemented. Id = NSMex02510 If 2 directories, both mountpoints, are specified for backup, and one mountpoint name is a subset of the other, one of the mountpoints is not backed up correctly. Resolution: Backup session manager now correctly handles nested mountpoint in host backup. Id = NSMbb22383 Package hostname should be added into the cell_info file in order to be able to backup shared objects on the relocatable hostname. MC/SG configuration doesn't add the relocatable hostname into the cell_info file, and in order for OmniBack II to function properly, the package hostname must be imported with 'additional hostname' manually. Resolution No fix in omnirforsg.ksh. Import must be done manually. See Admin Guide for more info on configuring the MS/SG integration. ID = NSMex01965 Due to invalid handling of "fork()" function, omnidbutil return code is "0" even if an error occurs. Part of SSPUX300_024. Resolution: Fork() handling improved. Id = NSMbb22689 Part of SSPUX300_029. Resolution: Id = NSMex01889 Session backup mode is not exported during the writeascii procedure. Because of this all the session modes are "Full" after the readascii. Part of SSPUX300_030. Resolution: Session mode is exported during the writeascii, and read back in during readascii. Id = NSMex01998 Wrong handling of device locking tables in bsm when dynamic device allocation is used. Part of SSPUX300_031. Resolution: Device locking handling improved. Id = NSMex02146 If the Library is configured for hostname longer than 32 characters, the error is reported: "Host for library LibraryName not found." Part of SSPUX300_045. Resolution: Hostname handling implemented for names longer than 32 characters. Id = NSMex02117 OmniBack II is not capable of decoding licenses containing '0' in the Cell Manager IP address. Part of SSPUX300_046. Resolution: Latest licensing API linked into crs daemon fixes decoding licenses created for IP addresses with '0' in it. Additionally, a change in license handling algorithm enables OmniBack II recognize and use such licenses correctly. Id = NSMex02009 Floating hostname in the mom_server file in MC/SG environment causes Cell Administration GUI to hang during installation of OB2 modules. Part of SSPUX300_052. Resolution: Hostname resolution fixed. Id = NSMex02242 In host backup only Mount point directories can be excluded. If a directory is not a mount point then it cannot be excluded from the backup. Part of SSPUX300_059. Resolution: The non-mountpoint directories can now be excluded. Id = NSMex02303 If a host backup object cannot be contacted in a multiple object backup specification, the backup exit status is 10 (Completed/Errors) instead of 11 (Completed/Failures). Part of SSPUX300_059. Resolution: Exit code in this situation is now 11. Id = NSMex02218 The backup hangs if one of the client systems is under heavy load and it can not send or receive the data promptly. Part of SSPUX300_059. Resolution: New mechamism implemented for aborting the session if the data are not sent or received within the timeout defined with "SmSendReceiveTimeout" global option. Id = HSLco14785 If a client fails with the message: "Could not connect to inet in order to start VBDA@hostname" then the object, started before the problematical one, fails with the error: "Got unexpected close from Unknown on." Part of SSPUX300_071. Resolution: Fix is implemented in bsm. Id = NSMex02351 The global option 'UpgradeIncrToFull' is not set - which means: 'Do not perform incremental backup if there is no full backup of the object'. Backup session manager claims the session will be aborted, but the backup is performed instead. Part of SSPUX300_071. Resolution: Bsm now handles UpgradeIncrToFull properly. PHSS_17299: Id = NSMbb21666 BSM doesn't handle unix disk agent '-tree ' option correctly if host backup is started, and the directory is not mountpoint on the unix disk agent system. As a result, the directory is not backed up. Part of: SSP30_03 Resolution: Fix in bsm. Id = NSMex01657 The restore session passes a large number of parameters to the inet service. However only 1024 are accepted. As a result, the Disk Agent restores less files than it should. Because additional parameters are needed for each file, the number of successfully restored files is approximately 900 only. Part of: N/A Resolution: Fix in bsm, rsm and inet. Id = HSLco11971 The global option DbFreeExtFileSpace is checked on Byte basis instead of MByte. This causes omnidbutil to not report the 'dbspace low' error when it should. Part of: N/A Resolution: Correct checking of free space in omnidbutil. Id = NSMex01913 'omniforsg.ksh -primary' script stops all OB2 daemons first, then it tries to run 'omnicc -import...' command to add a host into cell_info file. This fails because all the daemons are down at that time. Part of SSP30_14 Resolution: Fix in omniforsg.ksh script. Forking processes in crs is not handled correctly if a child process is started and not waited for conclusion. When crs starts processes in cluster aware OmniBack II installation, they remain as and take up space in process table. Eventually, this leads into system not being able to start any process until crs is shut down. Part of SSP30_22 Resolution: SIGCHLD set to SIG_DFL, if waitpid will be used (we want to wait for child to finish), or to SIG_IGN if we don't want to wait for the child. This makes sure no processes are left in the system. Id = HSLco12167 'omniforsg.ksh -secondary' script adds some trash into the cell_info file. Part of: N/A Resolution: Fix in omniforsg.ksh script. Id = NSMbb22116 At failover time the script csfailover.ksh starts OB2 daemons. After that, it starts the command 'omnidbutil -change_cell_name', which needs exclusive lock of the database, but it cannot be achieved, because crs has already restarted the backup sessions on the secondary system. The result is that the package is halted and no backups are done and all OB2 daemons are stopped. Part of SSP30_14. Resolution: Timeout added to enable cell name changing, and restart the sessions after the name has been changed. Id = NSMex01889 OmniBack does not export all the data during omnidbutil -writeascii. The data is not exported if sequence of levels of incremental backup is not strictly in ascending order. NOTE: The problem never happens if only FULL-INCR backup policy is used. Part of SSP30_15 Resolution: Fix in writeascii algorithm. Id = HSLco12023 Backup session exit status should be 11 (Completed/Failures) if one or more hosts are not reachable during backup. Currently it is 10 (Completed/errors). Part of: N/A Resolution: Fix in backup session status. Id = NSMex01884 CRS crashed for 2 reasons: if 2.55 client connected to it or due to malloc problem in stat function. Part of SSP30_11 Resolution: Fix in crs. Id = HSLco11955 Restore reports the message: "Medium not found." if it is started after the "List From Media" action. Resolution: Fix in shared library. Id = NSMex01965 Due to invalid handling of "fork" call, omnidbutil return code is "0" even if an error occured. Part of SSP30_24 Resolution: Fix in omnidbutil fork handling. SR: 5003450304 5003444786 5003461509 1653305110 Patch Files: /etc/opt/omni/sg/csfailover.ksh /opt/omni/lbin/bsm /opt/omni/lbin/crs /opt/omni/lbin/dbbda /opt/omni/lbin/dbsm /opt/omni/lbin/mmd /opt/omni/lbin/rsm /opt/omni/lib/libob2eadm.sl /opt/omni/lib/libob2ecmn.sl /opt/omni/lib/libob2emmdb.sl /opt/omni/newconfig/etc/opt/omni/sg/csfailover.ksh /opt/omni/sbin/install/omniforsg.ksh /opt/omni/sbin/omnidbutil /opt/omni/sbin/omnisv.sh what(1) Output: /etc/opt/omni/sg/csfailover.ksh: None /opt/omni/lbin/bsm: [TZ=GMT0 /opt/omni/lbin/bsm -version] HP OpenView OmniBack II A.03.00: BSM, internal build 1025, built on Fri Sep 17 16:10:50 1999 /opt/omni/lbin/crs: [TZ=GMT0 /opt/omni/lbin/crs -version] HP OpenView OmniBack II A.03.00: CRS, internal build 1025, built on Wed Jun 23 12:26:55 1999 /opt/omni/lbin/dbbda: [TZ=GMT0 /opt/omni/lbin/dbbda -version] HP OpenView OmniBack II A.03.00: DBBDA, internal bui ld 1025, built on Wed Aug 25 15:52:13 1999 /opt/omni/lbin/dbsm: [TZ=GMT0 /opt/omni/lbin/dbsm -version] HP OpenView OmniBack II A.03.00: DBSM, internal buil d 1025, built on Wed Jun 23 10:24:33 1999 /opt/omni/lbin/mmd: [TZ=GMT0 /opt/omni/lbin/mmd -version] HP OpenView OmniBack II A.03.00: MMD, internal build 1025, built on Wed Jun 23 12:54:46 1999 /opt/omni/lbin/rsm: [TZ=GMT0 /opt/omni/lbin/rsm -version] HP OpenView OmniBack II A.03.00: RSM, internal build 1025, built on Thu Jul 1 14:43:22 1999 /opt/omni/lib/libob2eadm.sl: None /opt/omni/lib/libob2ecmn.sl: None /opt/omni/lib/libob2emmdb.sl: None /opt/omni/newconfig/etc/opt/omni/sg/csfailover.ksh: None /opt/omni/sbin/install/omniforsg.ksh: None /opt/omni/sbin/omnidbutil: [TZ=GMT0 /opt/omni/sbin/omnidbutil -version] HP OpenView OmniBack II A.03.00: OMNIDBUTIL, interna l build 1025, built on Tue Oct 5 14:22:28 1 999 /opt/omni/sbin/omnisv.sh: None cksum(1) Output: 3236234904 2494 /etc/opt/omni/sg/csfailover.ksh 3461814734 1149863 /opt/omni/lbin/bsm 780600382 703332 /opt/omni/lbin/crs 4255105344 443724 /opt/omni/lbin/dbbda 2117734258 1212284 /opt/omni/lbin/dbsm 1567916083 522790 /opt/omni/lbin/mmd 3168634050 855819 /opt/omni/lbin/rsm 1947855315 651896 /opt/omni/lib/libob2eadm.sl 709106674 824689 /opt/omni/lib/libob2ecmn.sl 2257432585 648010 /opt/omni/lib/libob2emmdb.sl 3236234904 2494 /opt/omni/newconfig/etc/opt/omni/sg/ csfailover.ksh 1817396469 4794 /opt/omni/sbin/install/omniforsg.ksh 2301647157 656901 /opt/omni/sbin/omnidbutil 3863149994 11538 /opt/omni/sbin/omnisv.sh Patch Conflicts: None Patch Dependencies: None Hardware Dependencies: None Other Dependencies: None Supersedes: PHSS_17299 PHSS_18697 PHSS_19359 PHSS_19938 Equivalent Patches: PHSS_20084: s700: 11.00 s800: 11.00 Patch Package Size: 7590 KBytes Installation Instructions: Please review all instructions and the Hewlett-Packard SupportLine User Guide or your Hewlett-Packard support terms and conditions for precautions, scope of license, restrictions, and, limitation of liability and warranties, before installing this patch. ------------------------------------------------------------ 1. Back up your system before installing a patch. 2. Login as root. 3. Copy the patch to the /tmp directory. 4. Move to the /tmp directory and unshar the patch: cd /tmp sh PHSS_20083 5a. For a standalone system, run swinstall to install the patch: swinstall -x autoreboot=true -x match_target=true \ -s /tmp/PHSS_20083.depot By default swinstall will archive the original software in /var/adm/sw/patch/PHSS_20083. If you do not wish to retain a copy of the original software, you can create an empty file named /var/adm/sw/patch/PATCH_NOSAVE. WARNING: If this file exists when a patch is installed, the patch cannot be deinstalled. Please be careful when using this feature. It is recommended that you move the PHSS_20083.text file to /var/adm/sw/patch for future reference. To put this patch on a magnetic tape and install from the tape drive, use the command: dd if=/tmp/PHSS_20083.depot of=/dev/rmt/0m bs=2k Special Installation Instructions: PHSS_20083: NOTE! Patch has to be installed on the Cell Server host. "/opt/omni/sbin/omnisv.sh stop" must be executed on the cell server before swinstall is run. After the swinstall completes, restart OmniBackII using "/opt/omni/sbin/omnisv.sh start". NOTE! In order to resolve defects with Id NSMex01657 and NSMex02773, OmniBack II A.03.00 CORE patch (PHSS_20085/ PHSS_20086 or later) must also be installed and OmniBack II User Interface must be redistributed to the client systems. NOTE! A part of the NSMex02098 defect resolution is a new environment variable OB2DBRETRY. It is used to specify the number of retries before giving up to back up OB2 internal database. The variable should be set in the file /opt/omni/.omnirc on the Cell Manager system, using the following syntax: OB2DBRETRY=n 'n' is the number of retries before aborting the disk agent. NOTE: A selective write/readascii is introduced with the patch PHSS_18697/PHSS_18698. A list of hosts can be specified for which the processing should be done, instead of processing all the hosts in the cell. This is the procedure for enabling selective write/readascii: Selective writeascii 1. Start writeascii using the command 'omnidbutil -writeascii -cdb -mmdb ' 2. Stop the omnidbutil with Ctrl-C after it has reported "exporting filenames and fileversions for host ". 3. Edit the file hosts.txt, located in the directory from the omnidbutil command, and remove all the hosts except the ones that the writeascii is needed for. 4. Start the writeascii command again, with the option '-read_hosts'. Selective readascii 1. Edit the file hosts.txt and remove all the hosts except the ones that you want to perform readascii operation for. 2. Start readascii the usual way. It will read the hosts.txt file, and perform the readascii operation for the hosts listed there. Patch includes the following Site Specific Patches: SSPUX300_001 SSPUX300_011 SSPUX300_014 SSPUX300_015 SSPUX300_022 SSPUX300_024 SSPUX300_029 SSPUX300_030 SSPUX300_031 SSPUX300_045 SSPUX300_046 SSPUX300_052 SSPUX300_059 SSPUX300_071 SSPUX300_084