Parallel whole backups

[PeterP]

Not sure if this is an FYI or a request for help or what. But I discovered something interesting today.

Netbackup 6.0, IDS 11.70, Solaris.

onbar -b -w
BAR_MAX_BACKUP set to 10 or 0
This runs a parallel backup, i.e. one you need logs to restore.

onbar -b -w
BAR_MAX_BACKUP set to 1
This runs a serial 'whole' backup, no logs required.

So it seems the -w flag is redundant in some cases and if you specified this in your script it might be doing something other than you intend! Has this behaviour changed? I always thought it used to be a resource limit for onbar -b, NOT a secret config for whole backups.


Which asks the question, if usually I want to do a parallel backup with either a value for number of threads or 0 (unlimited) and once a week say do a 'whole' backup, how do I configure the onconfig?

Do I need to override the onconfig value with a setting/export in my whole netbackup script or pass in as a parameter or something?


As a side question, what's the best way to help netbackup.
1. netbackup deduplication (it works out todays backup was similar to yesterdays L0)
2. netbackup compression
3. Informix - BACKUP_FILTER of which I can choose: compress, gzip, bzip2, pbzip2, pigz?

Probably default values are fine:
BAR_NB_XPORT_COUNT - default 20
BAR_XFER_BUF_SIZE - default 31

TIA, PP.

[PeterP]

Google when it thinks I'm French finally found the answer.

https://www.ibm.com/developerworks/community/blogs/idsteam/entry/on_bar_whole_system_parallelize?lang=en

Historically people didn't use -w as it was slow compared to -b. So even if you are doing -w in parallel and have no logs you are fine.

Behaviour changed in cheetah, -w is now parallel if you want (once rootdbs is done). Quite why you would run a 'non' -w I don't know. It gives you the option to still restore if you have the logs or not.