NBRB parameters tuning

[Mohmec]

During the NBU configuration review for one of my customer I found the following “NetBackup resource broker” (NBRB) tuning parameters recommendations.

(1) Action to Consider: Execute command : nbrbutil -changesettings RB_RESPECT_REQUEST_PRIORITY=true
(2) Action to Consider: Execute command : nbrbutil –changesettings RB_BREAK_EVAL_ON_DEMAND=false

  If RESPECT_REQUEST_PRIORITY is set to true, then nbrb will stop, perform cleanup and restart the resource allocation evaluation cycle. This has the following advantages

• New requests are picked up.
• High priority jobs  (in the normal queue), which are at the top of the queue, gets a higher chance of running.

The report also mentions, “the tuning parameters is typically performed on medium to large NetBackup environments”.

Question: Is it a best practice to set evaluation cycle irrespective of size. What is considered as medium environment w.r.t nbrb. Any benchmark? Customer is backing about 50 TB using VADP. Can anyone let me know the settings in their environments (nbrbutil –listSettings)?

[Marianne]

You forgot to tell us your NBU version.

nbrb tuning has changed across versions.

[Mohmec]

NBU 7.6.0.1. New implementation.

[Marianne · Accepted solution]

Any reason why not 7.7.1 or 7.7.2? All versions up to 7.6.x will reach EOSL in a year from - Feb 2017. Those parameters are for environments that mainly use tape backups, lots of jobs and a variety of different priority settings in policies and Host Properties. The settings will determine whether higher priority will receive priority over drive/tape usage. Question therefore should be if this is important to the customer. In most environments these are left default. Especially modern setup that backs up to disk first.

[Mohmec]

Thank you Marianne.

Not sure why customer used 7.6. I was called in only to review the configuration.

[sdo]

Some take the view of only tuning when required, and not to attempt to pre-emptively pre-tune.  50TB for a full backup set isn't that large.

Why 7.6.0.1 for a new build, and not v7.7.2 ?

Haven't had to tune this on an N5230 v2.6.1.2 Appliance doing about 25TB weekly plain client, 380 clients, multi-streaming, so about 1000 jobs for the weekly session.  MSDP storage unit has 40 job limit.   Lots of jobs queue initially.  All policies use priority.  All schedules use an SLP.  Prod SLPs have an AIR replication stage, using priorities too.  Non-prod SLPs just have a backup stage.  No duplication to tape anywhere.  Haven't had to fiddle with RB tuning.

[Marianne]

Please tell them about 7.6 EOSL.