Vault or Live Cycle Policy to perform Duplications

[HEMANPR]

Hello People I will start to use a OpenStorage disk to perform my backups (Backup to Disk) but I have a question about do a duplicated... What is the best option to do a duplication? I use Vault Jobs right now to duplicated data into a tape, but i saw Live Cycle Policies can create a duplication of any policy without need to affect the network traffic. What is the best option Vaults or LCP???? Thanks

[HEMANPR]

Any Idea or Suggestion?

[AAlmroth]

Use NBU Vault if you like to control when to duplicate to tape/disk. NBU Vault does require that there is at least one tape library defined, but duplication targets can be disk.
Use SLP to automate the duplication fully. Vault relies on scheduled policies, whereas SLP kicks in when minimum threshold of amount of data is reached for a duplication session (tunable)

Vault +: You can control to duplicate when the backup system is idle.
Vault-: Expensive, but disk license to activate SLP is also fairly expensive

SLP+: Automatic once defined
SLP-: Occurs while there are other backup jobs running. Stressing disk systems with both read and write I/O tend to lead to poor performance.

/A

[Marianne]

Another ++ for SLP's:
Duplication is guaranteed - duplication is retried until successful.
SLP duplication can be suspended (from cmd) - useful when source media is needed for urgent restore. Vault duplication must be killed.

My main dislike in SLP is the fact that kick-off time cannot be scheduled/controlled.....

[David_McMullin · Accepted solution]

SLP is per policy or storage unit based, and allows multiple simultameous duplications unrelated to each other.
My experience with Vault is that the vault runs single threaded.
I had issues with how vault set up duplications, it would not use my drives efficiently, and would end up with one long running job when it could have run several shorter ones simultaneously.

[AAlmroth]

True, NBU Vault was designed when tape was the primary storage media. However, until SLP has matured, NBU Vault can be deployed instead when you need better control when to duplicate. In tiny to small environments (< 100 clients), SLP won't have the such a huge impact, but in any decent sized environment SLP is a pain point.

The common problem you mention is that duplication is single concurrent job per vault (other dup jobs stay queued). Each defined vault keeps a vault lock, and thus only one dup job can execute per vault. The trick around this is to define several vaults within a tape library definition. This way you can create several profiles split over many vaults, and thus you can utilise more drives. This approach does require more design work to balance the workload...

I'm eagerly waiting for scheduled SLP, or at least blackout windows...

/A

[AAlmroth]

The guarenteed policy is very nice, but it is not very easy to see queued jobs (not in GUI yet, perhaps 7.0.1?), but you have to go to CLI with the nbstlutil command.

The main problem I have experienced with installations, is that once an image has been queued for SLP duplication, it is "hard coded" to the destination. I've had numerous customers who can't understand why their second copy isn't available. And in 9 out of 10 cases, it is because they have removed the target storage unit (perhaps due to name change or decommision of media server). the only option is to actually cancel those image duplications. If the source image is still available, we could manually initiate a duplication once the queude item has been removed in SLP. Manual duplication doesn't seem to work when the item is still in the SLP queue.

I would like to see an option to nbstlutil where one could change the destination storage unit for already queued images...


And... scheduled SLP...

/A

[David_McMullin]

1. You can activate and deactivate SLP policies to schedule your duplications - which is unwieldy and a pain - but it works.

2. To list in process SLP jobs, I find the fastest way is this command:

bpimagelist -L -idonly -hoursago 72 -stl_incomplete

By default the hoursago is 24, I use aliases to also run this command periodically:

bpimagelist -L -idonly -hoursago 240 -stl_incomplete

If I have not duplicated in 10 days, there are major issues!

3. Since your SLP use the {install path}/netbackup/db/config/LIFECYCLE_PARAMETERS file, one thing I have noticed is that it has short and long retry periods - I set up a script that resets the jobs to the short period to force them.

# cat /var/adm/scripts/slpjob.activate.ksh
#!/usr/bin/ksh
#
# this lists each SLP job, sorts them by date and time, then activates each one
# which forces it into the short retry queue
# the last line starts a new scan of jobs
#
echo "Issuing command: '/usr/openv/netbackup/bin/admincmd/nbstlutil active  -wait -backupid $i' "
echo "For each backupid in '/usr/openv/netbackup/bin/admincmd/bpimagelist -L -idonly -hoursago 120 -stl_incomplete | sort +3 +4 +5n | /usr/bin/cut -f10'"
for i in `/usr/openv/netbackup/bin/admincmd/bpimagelist -L -idonly -hoursago 120 -stl_incomplete | sort +3 +4 +5n | /usr/bin/cut -d" " -f10`
do
 DATE=`date +%Y%m%d.%H%M`
 echo "Now activating "$i "at "$DATE
 /usr/openv/netbackup/bin/admincmd/nbstlutil active  -wait -backupid $i
 echo "------------------------------------------------"
done
/usr/openv/netbackup/bin/admincmd/nbstlutil new_session

[HEMANPR]

One more question
How can I add a catalog and Reports on a LCP?

Vault have those options easy to setup. LCP have those options?
Thanks

[Yogesh_Jadhav1]

Vault is best option if the backup window is during night hours. During day time we can schedule the vault job and set duplication jobs are per availability of drives. Also we can set the duplication run through SAN by selecting alternate read host as the media server on which robot is configured, if the environment is multi media servers and all connected through SAN.