Backupset generation incredibly slow

[krankin]

We know that we need to spread our database onto more spindles, but we've got some other issues, too.

We keep a copy stgpool of our main tape pool. When we generate a backupset, it seems to check every object in the tape stgpool and the offsite copy stgpool. Is there a way to avoid this in order to speed things up?

Any other backupset generation tips would be very welcomed, we're generating one right now for five average-size servers and it's been going for about 30 hours (and is only finishing up the third). This is on modern hardware, too.

[ttrinh7975]

krankin-

It's all depends on how much data are you generating the backupset from.
If the server have been around for awhile... The backupset will take time.

I ran into a backupset ran for 2 days. The node has about 250GB of storage
but over time those data grew...

Just gotta wait

[BBB]

>When we generate a backupset, it seems to check every object in the tape stgpool and
> the offsite copy stgpool. Is there a way to avoid this in order to speed things up?

As in it mounts all of the tapes?

Sounds like the node is not collocated, which could slow down the generate backupset by a great deal.

[piggyrat]

Depending the tape technology you are using a backup set generation is similar to a restore. If the data is a bunch of small files and any fragmentation on the tape (inactive object) then the process can be slow.

LTO is the worst with small files and other tape is not much better.

To resolve this issue:
1. Move all the node data to disk
2. Generate the backup set.

this will be a lot faster then a direct tape to tape generation.