|Article:TECH135484|||||Created: 2010-01-02|||||Updated: 2010-01-08|||||Article URL http://www.symantec.com/docs/TECH135484|
Below are the elements of NetBackup.
* Storage Units: Storage units are devices backups can be written to. There are two types of storage units - media manager units (tapes) and disk storage units. Operations on tapes - creating images, assigning media etc. are done via Enterprise Media Manager (EMM) and its Device and Volume Managers. Creating images on disks is done directly to disk storage. Drives of the same density must be in the same storage unit (e.g. two drives in one unit, with concurrent write set to 2).
* Storage Groups let you group units and set them as the destination in Policies.
* Policies: Policy type sets the type of clients that can be backed up by this policy (MS-SQL-Server types tells NetBackup that the policy contains only clients with the MSSQL option).
Related to policies:
Volume Pools: a group of media. Volume pool "NetBackup" is the default destination for backups other than the catalog backup. You can also create "Scratch pools" - for NetBackup to automatically transfer expired tapes (volumes) there. When another pool doesn't have any tapes with space available it will take empty tapes from the scratch pool. A key point: Tapes are returned to the Pool they were initially introduced into when all images have reached expiration. Meaning, if the tape was introduced into NetBackup into the NetBackup Volume Pool, it will return to the NetBackup Volume Pool, NOT the Scratch Pool. That being said, tapes should be initially introduced into the scratch pool so that they become available for use by any pool that requires media.
Checkpoints: Taking checkpoints will let you restart a job from the last checkpoint if it fails. Without checkpoints a restart will start from the beginning.
Job Priority: Higher value means higher priority. When netbackup has a free drive and a number of jobs waiting to be executed, it will take the job with higher priority first.
Compression: Is done on the client, before the data is sent to the backup server. It reduces the network load during backups. Your tape drive can also compress in its hardware. If that's the case then avoid using both tape and client compression at the same time, because it can actually increase the amount of data backed up. Only in rare cases is the client compression beneficial. In most cases it puts a heavy burden on the client machine's CPU.
True Image: This feature lets you restore data using one full backup and a number of incremental backups to the state it was in during the last incremental backup. Normally if a user created a folder or a file after a full backup and then deleted it after a couple of incremental backups you would still get them restored. With True Image, data that wasn't there on the last incremental backup would not be restored. Move detection: Incremental backups don't pick up files if their archive bit didn't change. When files or folders are moved the bit is unchanged. True image and move detection backups take more space and are slower
NB 6.5 implemented "data classification". It lets you send the backup data to a particular storage unit, storage lifecycle policy or storage unit group. You can create a "gold storage lifecycle" and only gold backup images will be permitted there.
Allow multiple data streams: backups can be divided into streams. If you have three slow drives in a server itsitï¿½s better to set them up in parallel, with three streams than one after another in one stream. You can also backup many drives from the same server in parallel if you have more than one tape drive. This way many streams of data are coming into the backup server. You should not use more than one stream per device (hard drive). Finding out the number of streams that will effectively run is tricky and depends on storage unit multiplexing, number of jobs that can start at the same time in a policy etc. Please also note that with this policy you will get as many jobs as there are streams in a backup. So if you are backing up three disks with three streams, you will get three jobs running and three backups from the same time period visible in the client if you want to run a restore.
Multiplexed jobs run as if they were separate inc or full jobs - for example backups with multiplexing enabled shows one job for each drive + one for the system state.
* Schedules: Calendar schedule lets you specify the run dates, weekdays, or days of the month for backups. Instead of setting schedule Frequency to 1 day or week you can set Calendar and configure specific days there.
* Backups types: Specified in schedule settings. An example would be if you retain differential-incremental backups for the same length of time as the full backups. The reasoning would be if a person created a file on Monday and deleted it on Tuesday the diff-inc backup would pick it up, while no full backup would contain it.
On Differential-incremental backups: Netbackup can determine file changes based on the archive bit or the date stamp. It is set in the client program. However, when files are copied from another computer, they keep their original timestamp.
* Frequency schedule settings (as opposed to the calendar settings): It tells Netbackup how much time has to elapse between runs of this particular backup. With this setting if you run a full backup on Monday, NetBackup will not try to start it until next Monday.
* Multiple Copies: Up to four real-time copies can be made of the same backed up data. If you have two drives, a job could be writing two copies at the same time, or a copy to disk and a copy to tape (each with different retention - e.g. keep the tape copy for three months and disk for one week).
* Override policy storage unit: you can run full backup scheduled jobs to one tape or disk, and all the other schedules to the policy storage unit.
* Retention: When the retention period expires for a tape, the data on the tapes can be retrieved only with the "import feature".
If you need day-to-day changes in data you must hold the incremental backups for as long as you do the full backups. Some folders may have different retention requirements, for them you can create separate policies.
You should specify retention periods longer than the run frequency (e.g. if you run every week, don't set the retention to one week but two). Archive schedules use infinite retention period. By default retention period of the first backup written to tape sets the ret period for that tape. You can't mix retention periods unless you use "allow multiple retention periods per media" option on the media server's properties, which is not recommended because you may have images with 1 week retention mixed with an image with an infinite retention period, therefore never expiring the images on that tape for use again.
There are two types of retention in NetBackup:
This determines how long you will retain the backup image on the tape.
Media (tape) retention:
This determines how long you can use the physical tape in NetBackup.
If you set media retention and the retention expires, NetBackup will never use that media again in its current state. It can use it for restores, but not for backups.
* Start Window: Here you specify the time window when backups in a policy can start. Be careful when using it with the calendar - If according to the calendar today is a run day and the start window covers it, backups will start (so if you have Sat and Sun set in the calendar and a window that covers Sat and Sun, the backup will run twice!).
If for example you have a backup window 11PM to 6AM and the calendar set to every day, jobs will run at 11PM, then again after 12AM, then again at 11PM of the same day etc. You can prevent that by either specifying start windows that don't span 12AM or leaving the same window, but setting the frequency of one day instead of the calendar (jobs set to run at frequency of one day will not be started again earlier than 24h after the last run).
One way of setting up the schedule for weekly backups is configuring frequency of one week with a start window every day 6:00PM to 8AM. This way you don't control what day the full backup runs - if it runs on Monday it won't try to run for 6 days. If it starts and fails on Monday, it will just retry on Tuesday during the run window. Next week it will run on Tuesday, not Monday.
Another way is running backups many times during a day. In this case you set it up to have a 24/7 start window and frequency of 12h.
* Backup Selections: The best practice when combining Windows 2000 and Windows 2003 clients in the same Policy is to use the System_State:\ or ALL_LOCAL_DRIVES directive. It is redundant to include ALL_LOCAL_DRIVES and/or System_State:\ and/or Shadow Copy Components:\. Please Note: As a best practice guideline, Symantec recommends one setting and not two or three in any combination.
Directive "NEW_STREAMS" is used to make the backup job start a new stream for whatever selections follow (ex. NEW_STREAM,file_1,file_2,NEW_STREAM,file_3 will give you two streams). Because you should not run more than one stream to a device (C: drive), this is not recommended.
* Synthetic Backups: With the traditional full backup all the files are copied from the client. Using synthetic backups netbackup merges an older full backup with what's on the server and creates a new full backup. The synthetic backups don't interact with the client, but are fully assembled on the server. The work best with backups on disk.
* NetBackup Catalog: The Catalog is an internal NB database that contains backup information (files and what media they are on) and NB configuration (storage devices, options etc.) NB needs this information to restore backups. Losing the catalog means losing the backups. The biggest part of the catalog is the Image Database (<install Path>\VERITAS\NetBackup\db\images for each server, with text files for each backup job and .f files with list of files backed up.
The next element is a relational DB (Sybase) in <install path>\VERITAS\NetBackupDB
* Catalog Backups: A catalog backup job starts three jobs 1) parent job 2) child job for the DB 3) child job for catalog images and configuration data. Both child job contain the actual data. To shrink the catalog size you can archive it (moving the large .f files to an archive). It requires a dedicated tape pool and a policy - once all is in place, the archive is started from the command line.
The small file that gets saved on the Master Server is the disaster recovery image file. It can also be sent to an emailed address.
The Catalog feature in the admin console lets you search backup images, duplicate them, promote images from a copy to the primary backup copy, expire images, import images from expired tapes.
Copies - each backup is assigned a primary copy, which is used for restores. If the primary copy ("original") is destroyed, you can promote a duplicate copy to a primary.
Article URL http://www.symantec.com/docs/TECH135484