HP Data Protector

by Mr bublers on 06/1/2012


Installation

How to migrate Fibre card from private-loop to fabric mode?

HP A5158A Fibre Channel Adapter Release Notes

Check Fibre Card:

ioscan -fnH will show detail of the device. The H/W path is also hardcoded on the back of the HP system.
fcmsutil /dev/td1 will actually check the fibre card status (Topology & Driver state).
Two most common FC in HP
A6795A HP Tachyon XL2 Fibre Channel Mass Storage Adapter
A5158A HP Tachyon TL/TS Fibre Channel Mass Storage Adapter

Check H/W Path: lssf /dev/rmt/0m

How to create Device file?

insf – used in general for creating, recreating device files on the fly. For eg: if you have added new devices, you can give insf command and it will in most cases, scan and create the corresponding device file (insf -C tape -e).
mksf – this is more generic. You can create a device file using mksf too, but here you need to give all the options and you can create one device file at a time.
mknod – You can create a device file through mknod too, but you need to have the exact major, minor number etc. It doesn’t have Class, driver, hw path and many other options available in the previous two commands to create a device file.

Was the hardware path with the FC interface used by another device before you connected the switch to it? In this case is a little possibility that another device file on this address blocks the creation of a new one. You can try to delete all devicefiles for this address (rmsf -H ) and create new one (infs -H -e)

OmniBack Processes:

Start/stop Omniback Service:
/sbin/rc2.d/S838omni –> essentially it runs /opt/omni/bin/omnisv.sh
3 main background processes:
/opt/omni/lbin/crs –> Cell Request Server
/opt/omni/lbin/mmd –> Media Management Daemon, started by crs
/opt/omni/lbin/rds -d –>

Manually install Solaris disk agent of Omniback

In client to be install: mkdir /tmp/omni_tmp; mkdir /tmp/omni_tmp/omnicf; mkdir /tmp/omni_tmp/da
FTP the following files:
backupserver:/opt/omni/databases/vendor/omnicf/sun/sparc/solaris-26/A.05.10/packet.Z –> /tmp/omni_tmp/omnicf
backupserver:/opt/omni/databases/vendor/da/sun/sparc/solaris-26/A.05.10/packet.Z –> /tmp/omni_tmp/da
backupserver:/opt/omni/databases/utils/sun/sparc/solaris-26/utils.tar –> /tmp/omni_tmp/
Install the core package & disk agent.
cd /tmp/omni_tmp/ && tar xvf utils.tar
Install the core package: ./omni_rinst.sh /tmp/omni_tmp/omnicf/packet.Z Core A.05.10 sun/sparc/solaris-26/A.05.10 /usr/omni psprdbk1.natappsunix.eng.domain.com
Install the Disk Agent: ./omni_rinst.sh /tmp/omni_tmp/da/packet.Z DiskAgent A.05.10 sun/sparc/solaris-26/A.05.10 /usr/omni psprdbk1.natappsunix.eng.domain.com
Modify the configuration files.
Modify /etc/services with the line: omni 5555/tcp # DATA-PROTECTOR
Make sure the /etc/inetd.conf has the line: omni stream tcp nowait root /opt/omni//lbin/inet inet -log /var/opt/omni//log/inet.log
Restart the inetd (or kill -HUP) and telnet to port 5555 to make sure getting response.
Using Omni GUI, import the client.

How to install Oracle 8/9i Omniback Agent?

Shutdown the database;
Push the database agent;
For Oracle 8/9i, relink the dp database library with Oracle binary:
Solaris
cd $ORACLE_HOME/lib
for 64bit, mv libobk.so libobk.so.orig
ln -s /opt/omni/lib/libob2oracle8_64bit.so libobk.so
HPUX
Change to the /lib directory:
cd /lib (32-bit Oracle) or cd /lib64 (64-bit Oracle8i)
cd /lib (64-bit Oracle9i).
mv libobk.sl libobk.sl.orig
ln -s /opt/omni/lib/libob2oracle8.sl libobk.sl (32-bit Oracle) or ln -s /opt/omni/lib/libob2oracle8_64bit.sl libobk.sl (64-bit Oracle)
Startup Oracle instance.

Operation

How to migrate cell server to another host?

Backup all Omni related files just in case. Many prefer to put things out to tape..or you can cheat like me and just do an omni_write to another seperate mountpoint. I also copy down /etc/opt/omni and /var/opt/omni….just in case, under that new mountpoint too.
Load all your software on the new cell server (you can do this by ‘pushing’ out the software from your OLD cell server or by loading it via the CD).
Copy /etc/opt/omni and /var/opt/omni from OLD to NEW. (remember you have a backup copy of your db on that seperate mountpoint on your old system, just in case you need to rebuild and you have a tape if you did one-so your covered double).
On NEW: Edit /etc/opt/omni/cell/cell_server file to reflect NEW server_name
On OLD: Stop OBII
/opt/omni/sbin/omnisv.sh stop
On NEW: Start OBII
/opt/omni/sbin/omnisv.sh start
Chg Cell_Mgr_Name
/opt/omni/sbin/omnidbutil -change_cell_name
Check name change
/opt/omni/sbin/omnidbutil -show_cell_name
On All other cells:
Edit cell_server files to new cell_mgr name
/etc/opt/omni/cell_server
Make sure your NEW server controls your robotic arm. You will probably need to move the scsi connection so your other (NEW) server has it. And of course, you must be sure your kernel is configured for this with the appropriate driver. Using ‘uma’ to ensure you can move the tapes in and out to confirm the new server is controlling the arm.
Go into your OBII and under Devices/Media, highlight the Library and go out and modify in there so your NEW host is controlling the arm in the software too.
Now make sure your OBII is running correctly on your new server. Does it come up OK. Do you see your old sessions. Everything should be there because you copied the only directories you needed /var/opt/omni and /etc/opt/omni…Try to create a dummy backup on something small (not a fs on OLD host..) on one of the other clients, make sure it works.

How to configuring SNMP Reporting in Omniback?

Under /etc/opt/omni/snmp, there are 2 files:
OVdests –> dest of snmp trap messages
trap-dest: psdevov1.natappsunix.eng.domain.com
OVfilter –> message level to be filter out
warning
minor

How to add User to access Omniback?

/etc/opt/omni/users/UserList
eg, add “” “NSU” “*” * “admin”

How to check drive index?

From the system with Robot Arm connected (SCSI), locate device.

/opt/omni/lbin/uma -ioctl /dev/rac/c2t6d0
stat d –> list drive info
stat s –> list slot info
move s1 d1 –> moving tape from slot 1 to drive 1
move d1 d2 –> moving tape from drive 1 to drive 2
quit

From the system with Fibre Channel connected (Fibre), do:

mt -t /dev/rmt/1m offl –> checking drive 1 is really the 1m

Troubleshooting:

Error: [80:1003] Can not allocate/attach shared memory (IPC Can not Create Shared Memory Segment System error: [22] Invalid argument) => aborting.
Increase shared memory in client system (for solairs, increase SHMMAX to minimal 128M-129536000).
Offical Post from HP ITRC:
Cause: SCO OpenServer has system-imposed maximum shared memory segment size, defined by the kernel parameter SHMMAX with default value 512KB.
Solution: The size of the shared memory segment allocated by OmniBack II is determined as a product of the device block size (defined in .omnirc variable OB2BLKSIZE, with a default of 64K) and the number of buffers per Disk Agent, defined when configuring a logical device (with a default of 8).
Error Message: Unexpected close reading NET message => aborting
Cause: OmniBack II tried to allocate a larger shared memory segment and failed.
Solution: Either decrease the block size and/or number of buffers per Disk Agent (which also decreases backup performance), or reconfigure the kernel parameter SHMMAX to a suggested 2MB.
Error: Can not connect to Media Agent on system tor3ghp1.domain.com, port 49232 (IPC Invalid Hostname or IP Address System error: HOST_NOT_FOUND ) => aborting.
This is a name resolution problem, most likely to do with DNS. Omniback seems to parse result of nslookup for correct hostname. In Solaris, nslookup will not use resolver library, which means it doesn’t care anything of /etc/nsswitch.conf. Nslookup in Solaris will not use /etc/hosts file as well (same as standard bind’s nslookup). Nslookup in HPUX is behaving different and it uses /etc/hosts file based on /etc/nsswitch.conf.
In situation where changing /etc/resolv.conf is NOT an option, I have made a simple script which will emulate nslookup and format a similar output, but it checks /etc/hosts based on /etc/nsswitch.conf. It works fine with Omniback in Solaris clients. Download the script from here.
Error: Omniback Database 8/9i Integration “Could not Export Recovery Catalog” at the end of database backup.
Cause: Target database’s version doesn’t match with Rman database and its exp utility could not export Recovery Catalog properly.
Solution:
Find out the Oracle Version of Rman catalog database (just run sqlplus), such as “Release 9.2.0.1.0 – 64bit Production”.
Find out the Oracle version of target database, such as “Release 9.2.0.6.0 – Production”.
FTP Rman database’s exp binary from $ORACLE_HOME/bin/exp to Target Database’s $ORACLE_HOME/bin/exp.rman. Check to permission to be 751.
In Omniback backup specification, open “Source” tab, right click the database name and choose “Set Environment Variables”. Set “ORA_EXP” to be Target Database’s $ORACLE_HOME/bin/exp.rman.
Error: RMAN-06004: ORACLE error from recovery catalog database: RMAN-20003: target database incarnation not found in recovery catalog.
Cause: The target database has been changes (SCN, etc.). eg: target database was opened with the resetlogs option
Solution: Use RESET DATABASE to register the new version.
Error: “Oracle8 is not linked with the Data Protector library for backup”.
Solution: Most likely is the Rman catalog listener down.

Database

How to remove old sessions and decrease the size of your database in Omni3.5?

Firstly, the purge session is scheduled to run every day at 12:00. This is the omniback default setting. This can be done manually as:

omnidb -strip
omnidbutil -purge -dcbf
omnidbutil -purge -filenames
omnidbutil -purge -messages 60
omnidbutil -purge -sessions 60
omnidbutil -purge -days 60

If you still have session showing since April they can be removed manually. This is done by viewing previous sessions and then selecting Actions -> Remove session.
Information about a session remains in the Database until the catalog protection expires, then it is automaticlaly removed by a purge session. You may want to check what catalog protection you have set for each session/backup.

To reduce the size of your database (and hence file versions) in version 3.5 you can do this:

omnidbutil -extendinfo (this gives size of fvers and fnames)
stop and then start omniback daemons
/opt/omni/sbin/omnisv.sh -stop
/opt/omni/sbin/omnisv.sh -start
backup your database and omniback areas (/var/opt/omni, /etc/opt/omni, /opt/omni)
create a temp directory for mmdb and cdb
e.g. /tmp/cdb and /tmp/mmdb where ever you have enough space.
omnidbutil -writeascii -mmdb /tmp/mmdb -cdb /tmp/cdb (this writes the database to an ascii file)
omnidbutil -readascii -mmdb /tmp/mmdb -cdb /tmp/cdb (this reads database back).
omnidbutil -extendinfo for new sizes of fnames and fvers.

Commands

How to start Omniback backup from command line?

echo “Starting Backup `date ‘+%m/%d/%y – %H:%M%’`”
/opt/omni/bin/omnib -datalist BackupMain_LTO

Main Reporting Tool: /opt/omni/bin/omnirpt

/opt/omni/sbin/utilns/get_info # generate /tmp/get_info.rpt file contains all omniback system information
omnidb
omnidb -strip # If no option is specified, it strips catalogs on all data objects that are no longer protected.
/opt/omni/bin/omnidb -session -since 2004/12/16 -until 2004/12/27 -detail # session information from 04/12/16 to 04/12/27
omnidb -session -latest -detail # latest session information
omnidb -session -last 3 # backup stations of last 3 days
/opt/omni/bin/omnidb -filesystem :

General

Which backup programs should I choose?

Using SAM to backup (or better yet, the command line for fbackup) works just fine and has huge advantages over tar, cpio, and pax:
high speed restore (a few minutes to get to the last file)
Centralized index so you don’t have to spend an hour reading the entire tape to see what is on the tape
multi-tape indexes that track which files are on which reel of a multi-tape backup.
error recovery and resync for tape problems
supports large files
can use up to 6 slave programs to keep the tape busy and avoid data underruns
Use tar for small file data interchange (for which tar was designed).

How to write-protect the DDS tapes?

Always check the Write-Protect tab before doing backup!!!
You can write-protect or write-enable a DAT tape by using the sliding write-protect tab. You can only write data to the tape when the tab is in the closed position.

How to manuplate tape devices?

Search for tape devices: ioscan -kfn -C tape
List detail of tape devices: lssf /dev/rmt/0m
Rewind tape devices: mt -f /dev/rmt/0mnb rew

How to read tape?

By tar: tar tvf /dev/rmt/0m
By fbackup: frecover -I /tmp/index -f /dev/rmt/0m
This will give a index-file in the file /tmp/index, if the tape-device is /dev/rmt/0m
By cpio: cpio -ictB All others: od -c

How to backup to remote tape drive?

Using GNUtar 1.13.19 above is one of the best options. HPUX default tar can not understand hostname:/dev/rmt/0m format.
Backup to tape: gtar cvf :/dev/rmt/0m


Append to tape: gtar rvf :/dev/rmt/0m

How to backup with compression?

If the tape drive support ‘hardware compress’ (such as what I have: C1537A DDS-3), create a device file in /dev/rmt with compression.
System Default Compress Tape Devices might NOT work ideally. These devices are:
lssf /dev/rmt/* | grep ‘best density available’
If the tape drive doesn’t support ‘hardware compress’, then use software compress. Something like below will do the job.
tar cf – . | gzip | dd of=/dev/rmt/0m obs=100k
For fbackup, just use /dev/rmt/0m if lssf /dev/rmt/0m and ioscan -knfC tape reports that this is your tape drive. The 0m file will enable data compression and fbackup will handle the tape positioning.

Aku tulis sendiri semua ni ? jangan harap .. ni aku jumpak document orang …
Regards,
/bublers

There are 2 comments in this article:

  1. 12/8/2012Fuad Baharuddin says:

    hi bro,

    nice posting and info.

    Nak tanya skit, am running HP Data Protector version 6.00 on Windows platform. I could backup the servers mostly HPUX very fast via FC. Unfortunately, when restoring time it’s taken longer than a backup time. How to set/ensure the restoration is using FC instead of network TCP.

  2. 01/21/2013Mr bublers says:

    Use the same drive yang digunakan untuk backup (yang zone(d) ke server)

Write a comment:

You have to log in to write a comment.

More in Application, bubmy (271 of 481 articles)