NSR Not Using All Drives in JB 1

[Arluin]

I'm running networker 6.1.3 on Sol8 with a Sun L100 jb (Quantum ATL robotics, 3 HP LTO2 drives).

During backups of my 5 groups/1 pool the jb will load tapes in all three drives, but only two of the tapes ever get written to.

I've set the device target sessions to 2 to try to limit sessions on the two drives so that it is forced to use the 3rd drive.

Any ideas of how to spread the load so that I can reduce the overall backup time (full backups take 10 hours for 712GB).

100baseT over switched network.

 

[605]

Make sure that the server parallelism has been set high enough - the default value is 4.

 

[Arluin]

I wasn't sure if parallelism was relevant so I left it off the initial post. I raised it to 6 and reset the jb and rebooted the storageserver. I then raised it to 10 but didn't reset/reboot. I rarely get more than 4 sessions going at a time and I don't know where the excess sessions go, but it's not to the 3rd drive.

 

[605]

Also make sure that you have your pool setup either
- for all drives or
- for no drives

Although this parameter is rarely feasible with respect to
technical reasons, customers love to check the devices within
then pool's setup. As a consequence, not all drives might be used.

Also look that the third drive is not also set for "read only".

BTW - you must not reset/reboot your system to make changes become effective.

Also, if you leave the save set entry for the client to "ALL",
there is only a new save set for each partition. So if you
need more save sets, you have to specify them manually.

 

[Arluin]

Missed a 'parallelism' setting. I set the one under Media | Jukebox, but there is another under Server. That was still at 4. Once I throttled it up, I was pushing 3 sessions each to all three drives.

Thanks for the help.

 

[wallace88]

After increasing the server parallelism (and therefore the number of savestreams that are simultaneously backed up onto your backup devices), I would strongly advise you to perform a test recovery to see what type of recover performance you will experience.

The server parallelism can be viewed as a pivoting point. Setting it higher can lower the time it takes to backup up data, but recoveries will take longer. And vice versa.

Some users set this to the maximum, and then discover that it takes a long time to recover files.

I usually recommend that a user determine what your business will allow for the maximum amount of time to recover data, benchmark the recovery process, then modify the environment so that you can achieve this goal. Modifications can be in the form of (to name a few...):

- adjusting server/client parallelism
- adding more tape drives
- dedicated backup network
- backing up to different pools
- using storage nodes

You can also look at the following document for additional informaiton:

Legato NetWorker Cross-Platform Version Performance Tuning Guide, Release 7.1

http://web1.legato.com/infodev/publications/NetWorker/perftun/7.1/perftune_7.1.pdf