San Francisco

dave spink toolset




CATALOGS

Master Server Catalogs

Directory Purpose
/usr/openv/netbackup/db/class Policy catalog
/usr/openv/netbackup/db/config Global settings, storage unit catalog
/usr/openv/netbackup/db/client Client specific e.g. DMZ clients here
/usr/openv/netbackup/db/error All errors, daily errors in failure_history
/usr/openv/netbackup/db/jobs Info on jobs
/usr/openv/netbackup/db/images Info of files on images "/fileX, date, image(9050..), media server"
/usr/openv/volmgr/database (volume cat) Volume catalog "volume id (A00001), robot 0 and slot 23"

Media Server Catalogs

Directory Purpose
/usr/openv/netbackup/db/media Info on media used "image(9050..), volume id (A00001)"
/usr/openv/volmgr/database Device catalog for info on drives and robots.

See an example of this relationship between a Netbackup Master and Media Server during a "restore process":

  1. Via restore GUI screen - specify Master Server, Source Client and Destination Client.
  2. Master server contains the images catalog, and allows you to browse files for restore.
  3. Master server contacts media server "that wrote the backup to tape" for volume id.
  4. Media server reads media catalog and returns volume id and its status to master server.
  5. Master server reads volume catalog (volDB) for tape / slot number and loads tape or requests tape to be loaded.
  6. Master server scheduler contacts media server restore manager to handle request.
  7. Media server goes through process of contacting client, positioning tape etc for restore.

Common Netbackup Ports

bprd    	13720/tcp
bpdbm   	13721/tcp
bpjava-msvc	13722/tcp
bpjobd  	13723/tcp
vnetd   	13724/tcp
bpcd    	13782/tcp
vopied  	13783/tcp
nbdbd   	13784/tcp
visd    	9284/tcp

Media Manager Ports.

vmd     	13701/tcp
tldcd   	13711/tcp


DAEMONS

Master Server.

Daemon Purpose
bprd Responds to requests, starts bpdbm, calls bpsched (every 10 min.)
bpsched Scheduler (checks policy for backups due), calls bpdbm & media bpcd
bpdbm Manages catalogs, check clientname with list, calls bpjobd
bpjobd Job monitor (feeds activity monitor), calls nothing


Media Server.

Daemon Purpose
inetd Listener, started by bpsched, calls bpcd.
bpcd Accepts incoming netbackup requests, calls bpbrm. May run as "-standalone" mode.
bpbrm Backup and restore mngr, calls bptm/bpdm & bpcd (client).
bptm (parent) Takes shared memory image from bptm (client) and stores on tape.
bptm (client) Receives data from client bpbkar process, stores image in memory.
bpdm (parent) Takes shared memory image from bpdm (client) and stores on disk.
bpdm (client) Receives data from client bpbkar process, stores image in memory.


Clients.

Daemons Purpose
inetd Listener called by bpbrm, calls client bpcd.
bpcd Accepts incoming netbackup requests, calls bpbkar.
bpbkar Backup manager, sends image to bptm/bpdm, sends metadata to master via bpbrm.


Media Management Daemons.

Daemons Purpose
ltid (logical tape interface daemon) Handles TLU. Calls "vmd" for media info.
vmd (media manager daemon) Updates volumes catalog, responds to ltid.
txxd (ours is tldd) Handles drive requests t l-library d-DLT d-daemon.
txxcd (ours is tldcd) Handles arm requests.
avrd Reads tape labels

Note, to bypass inetd you can run "bpcd -standalone", for this to work add "ALLOW_NON_RESERVED_PORTS" to bp.conf.



MEDIA

The Media ID uses an external volume serial number (evsn) and a recorded volume serial number (rvsn). The external bar code reads 13 characters while internal bar code uses last 6 of 13 characters. You specify which 6 of 13 chars to use for Media ID in vm.conf. When a volume is mounted for first time the "rsvn" is written & "time assignment" updated.

Volume Pools are used to separate data i.e. different volume pools cannot use the same tapes. The defaults pools are - netbackup (for catalogs), none (cleaning tapes), and datastore (third party). A scatch pool exists that moves volumes (tapes) from scratch to another pool as needed.

Volume Groups are used to ease management of tapes. For example, "00_000_TLD" robotic group has our LTO tapes and "01_001_TLD" robotic group has our DLT tapes. The "---" is named given to standalone tapes i.e. ones removed from the robots. The media server "DEFAULT" when importing tapes means it checks volume groups and places compatible tapes into an existing group.

Delete Volumes can only occur when no images exist i.e. an assigned volume (contains images) cannot be deleted. To deassign a volume run "bpexpdate -m mediaID -d 0". The "time assigned" column is quick indicator of netbackup storing images on a tape. When all images have expired "time assigned" is cleared, as with entries in catalogs. There are two expiration types; volume expiration (how often tape used) and image expiration (retention time of data).



POLICIES

The due diligence is to have one data stream per physical spindle. For Unix clients with multiple file systems be sure not to select "cross mount points" with ALL_LOCAL_DRIVES directive. Assume unix system has 5 mount points, Netbackup queues 5 streams regardless of whether "cross mount points" is selected. Essentially the ALL_LOCAL_DRIVES directive ensures all unix file systems form their own backup stream.

There are times when a mount point contains too many files and you need to create more streams for that file system. To manually configure multiple data streams you can create a file list and ensure "Allow multiple data streams" is selected in policy.

NEW_STREAM
/usr
/lib
NEW_STREAM
/home
/bin 

Exclude / Include list locations for Unix. Windows is modified via GUI; client properties.

/usr/openv/netbackup/exclude_list or include_list
/usr/openv/netbackup/exclude_list.policy_name

A User-Directed backups require a schedule called "user backup" with a backup window defined. The client must also reside in the policy. The user selects files via their supplied list and hence netbackup ignores files in policy.



OTHER

Netbackup does not mix retention periods on a single volume. The images entries are deleted when the retention period ends. The expired image becomes invalid after midnight, while the image metadata is deleted from images catalog when image cleanup is performed. Once an image expires the only way to restore is to re-import tape. See Master Server properties retention periods.

SSO - Shared Storage Option allows sharing of tapes drives across multiple media servers. Two media servers cannot write to the same tape drive at the same time. A media server performs "scsi reserve" on the drive until it has completed it's backup. To configure a robot for shared drives the robot must first be zoned to all media servers. If robotic control is SCSI-based only one media server controls robot tape loading and unloading, while all media servers use the robot for backup and restores.

SAN Media Servers - basically clients with assigned tapes drives i.e. they can only backup themselves.

Storage Units - are logical devices that NetBackup uses to write backups. A media server sees a Storage Unit as a device to store backups. When creating a Storage Unit you specify the tape library to use, fragment size, numbers of concurrent drives, and multiplexing number.

Maximum Fragment Size - Lets assume you had 70GB backup and needed to restore file located at 69GB. With fragment size value of 2G netbackup skips 34 file markers and starts read at 68GB. Hence fragmentation is used to enable fast location of file or directory.

Multistreaming - Allows a single clients to send multiple streams. The best practice is one stream per disk to avoid "disk thrashing".