NB4.5 Catalog Backup to Tape

[104addo]

Hi

Currently we have a Netbackup 4.5 server (Solaris Platform). I've configured the Catalog to back up to disk one evening and to tape another evening. The Catalog backup to tape appears to "hang" and displays as still ACTIVE after my backups have finished. The catalog size is approx 650 mb trying to backup onto DLT 40/80 gb tapes...

I've tried a manual backup using bpbackupdb -m xxxxx to no avail.

Any help would be greatly appreciated.

 

[Stumpr]

3 items

1)
Under Solaris operating systems, the catalog backup is very slow.

http://seer.support.veritas.com/docs/273288.htm

Details:
A catalog backup running on a Solaris server may take an extraordinary amount of time, far longer than normal file system backups of the same data. This happens especially when the server has been configured for Sun Trunking.

The problem is two TCP values that can get set quite high on a Solaris servers: tcp_xmit_hiwat and tcp_recv_hiwat. Run the following commands to see what their current values are:

ndd -get /dev/tcp tcp_xmit_hiwat
ndd -get /dev/tcp tcp_xmit_hiwat

If these values are set to 65536 or higher, the catalog backup may run very slowly.

Currently, the only solution is to drop those values down to 49152. Execute the following commands:

ndd -set /dev/tcp tcp_xmit_hiwat 49152
ndd -set /dev/tcp tcp_recv_hiwat 49152

That will change those values until the server is rebooted. Also, add the above two lines to the server's S69inet script in the /etc/rc2.d directory.

After making this change, the catalog backup performance should improve dramatically.

Note: In the event that trunking is desired, VERITAS has been working with Sun on this issue. The install recommendations for Sun Trunking specifies that tcp_recv_hiwat and tcp_xmit_hiwat be set to 65536. However, Sun says that although this is the recommended setting, the value of 49152 will also work with Trunking without any problems.


2)
dbbackup_notify returns tape device path, when bpbackupdb is run with -tpath option

http://seer.support.veritas.com/docs/247000.htm

Details:
When the NetBackup DB backup is called from the command line with the following arguments -

/usr/openv/netbackup/bin/admincmd/bpbackupdb -tpath tape_device_path [ -rv recorded_vsn ] [ -nodbpaths ] [ -v ] [ path ... ]

NetBackup passes the following parameters to the dbbackup_notify script:
Parameter Description
device Device type the backup was written to. In this case it will be TAPE
tape_device_path Device path to the tape drive, where the tape is mounted.
status Specifies whether the backup was successful and must have a value of either SUCCESS or FAIL.


For example:
dbbackup_notify TAPE /dev/rmt0 SUCCESS


Note: Even if the -rv option is used with the bpbackupdb command, the tape_device_path parameter is used, and not the recorded_vsn.

If for scripting purposes a recorded_vsn is needed, mount a tape (tpreq command) under a device path with the name that would contain the media ID used,

For example:
dbbackup_notify TAPE /tmp/sync_DLT001 SUCCESS

Then, parse the media ID out of the tape_device_path parameter.


What is the dbbackup_notify script?

The dbbackup_notify script runs on the master server and is called each time VERITAS NetBackup completes an attempt to back up its catalogs.

The dbbackup_notify script is described in more detail in the VERITAS NetBackup System Administrator's Guide.


3)
Scheduled backup doesn't run and there are no active or queued jobs in the Activity Monitor.

http://seer.support.veritas.com/docs/234894.htm

Details:
One cause of this behavior is a hung catalog backup. Scheduled backups will not run while a catalog backup is running. Check if a catalog backup is running by using the /usr/openv/NetBackup/bin/bpps -a command. If a catalog backup is running a line under NetBackup Processes group will show a bpbackupdb if the catalog backup is running. If a catalog backup is running longer than usual, look for the following to determine if there is a media problem.

The catalog backup may be hung because the media assigned for the backup is bad or unavailable. Check the Catalog backup configuration to see what media is assigned. Verify that the media is in the robot and has the correct robot type, number and slot number assigned. Also check the Device Management screen for pending requests. If this looks fine, the tape may be bad. In this case change the media assigned to do the catalog backup. After changing the media assigned to the catalog backup, shutdown the NetBackup processes to kill the catalog backup. Restart the NetBackup processes and start a catalog backup. Refer to page 146 in the NetBackup DataCenter System Administrator's Guide for information on changing the configuration of your catalog backups. If using NetBackup 3.2, refer to page 250 in the NetBackup for Unix System Administrator's Guide.


Well, thats a start
BTW
only 650 MB? ...That is way small.
Are you doing the images as a separate backup?
If not then why is the catalog backup so small?



Bob Stump
Incorrigible punster -- Do not incorrige

 

[104addo]

Thanks for the info. I shall go through these and I'll check out the size of the Catalog as well!!!!

 

[104addo]

Hi Stumpr and other interested parties!

I tried your suggestions 1 and 2. The settings for the tcp_xmit_hiwat were set at 24576 so there appeared to be no issue there. I did suggestion 2 and I got back successful results.

I created an admin directory under /usr/openv/netbackup/logs and re-ran the catalog backup to tape. In the log it was trying to connect to an "old" server name which was incorrect. The disk catalog backup which works fine points to the correct server name.

therefore, how on earth do I configure/point that tape catalog backup to my correct server name. the bp.conf is correct...

Any help again would be great!!

 

[Stumpr]

you can do this via the GUI at
"Catalog" ->
"Configure NetBackup Catalog backup"On the "Attribute" tab
enter the correct servername in the field called media server.

you can also do it via the command line using bpsyncinfo
# bpsyncinfo -help
USAGE: bpsyncinfo [-l|-L|-U] [-v] [-M master_server,...]
bpsyncinfo -delete <id> [-v] [-M master_server,...]
bpsyncinfo -doBackup [-v] [-M master_server,...]
bpsyncinfo: [-paths <"path1 path2 ...">] [-v] [-M master_server,...]
[-server <hostname connected to device>]
[-bw <backup when>]
[-id1 <id> [-d1 <densitystr>]]
[-id2 <id> [-d2 <densitystr>]]

densitystr is required if media type is tape
the list of paths must be enclosed in double quotes and blank-delimited
backup when choices: never, session or dbupdate
id is either a media ID or path

Bob Stump
Incorrigible punster -- Do not incorrige