Label Libvol Behaviour

JoseOlivenca

Newcomer
Joined
May 23, 2016
Messages
4
Reaction score
0
Points
0
PREDATAR Control23

Hi.

I'm posting this question here because I'm using a VTL but this can be a more generic question as to how the Label Libvol works.

Setup
- TSM: 7.1.3
- OS: Win 2012 R2
- VTL: Fujitsu EternusCS 8400
- VTL Robotics Emulation: ACSLS
- VTL Drive Emulation: 60 * IBM LTO4

The VTL (just because this isn't a very common VTL):
This is a Backup to Disk to Tape system. The VTL has small amount of cache (about 60TB) for the day to day use and a large tape pool (500TB at the moment but will be expanded in the short term).

The Problem:
I have a script that runs every day that labels all the medias that where deleted from a storage pool in the previous day. The ACSLS library type doesn't have the RELABELSCRATCH option.
As only the most recent volumes are kept in the disk cache when I issue the Label Libvol command most of the time the volumes are on tape.
When TSM requests a volume that exists only on tape it will have to be retrieved from tape to disk and then be made available to TSM. In this example it took 4 minutes to do a Label Libvol.

05/23/2016 16:13:54 ANR2017I Administrator ADMIN issued command: LABEL LIBVOLUME acslib G04089 overwrite=yes waitt=0 (SESSION: 298428)
05/23/2016 16:13:54 ANR0984I Process 1111 for LABEL LIBVOLUME started in the BACKGROUND at 16:13:54. (SESSION: 298428, PROCESS: 1111)
05/23/2016 16:13:54 ANR8799I LABEL LIBVOLUME: Operation for library ACSLIB started as process 1111. (SESSION: 298428, PROCESS: 1111)
05/23/2016 16:13:54 ANR0405I Session 298428 ended for administrator ADMIN (WinNT). (SESSION: 298428)
05/23/2016 16:17:52 ANR8810I Volume G04089 has been labeled in library ACSLIB. (SESSION: 298428, PROCESS: 1111)
05/23/2016 16:17:53 ANR0407I Session 298495 started for administrator ADMIN (WinNT) (Tcp/Ip ). (SESSION: 298495)
05/23/2016 16:17:54 ANR8800I LABEL LIBVOLUME for volume G04089 in library ACSLIB completed successfully. (SESSION: 298428, PROCESS: 1111)
05/23/2016 16:17:54 ANR0985I Process 1111 for LABEL LIBVOLUME running in the BACKGROUND completed with completion state SUCCESS at 16:17:54. (SESSION: 298428, PROCESS: 1111)

This morning I had 58 medias to label. The script issued the commands and all looks well. Except it wasn't.
The medias are labeled sequentially, one at a time. No media operations are performed during this time, no tapes are dismounted (a query mount shows medias as "dismounting") no tapes are mounted (there are MediaW on the clients that started a backup during this period). Even the cancel process get a cancel pending.
During this process there are several unused drives for TSM (30+) to use.

The question:
Is this the "normal" way the Label Libvol works?
Is there a way to parallel several label libvol processes?
Is there a way to minimize the priority for the label libvol processes?

Thanks in advance,
J.Olivenca
 
PREDATAR Control23

One thing that strikes me - why do you have to do this - "... labels all the medias that where deleted from a storage pool in the previous day"?

Why are the media deleted? If the VTs (Virtual Tapes) are empty, these should return to scratch (assuming reuse delay is set to ZERO) and you don't have to relabel.
 
PREDATAR Control23

One thing that strikes me - why do you have to do this - "... labels all the medias that where deleted from a storage pool in the previous day"?

Why are the media deleted? If the VTs (Virtual Tapes) are empty, these should return to scratch (assuming reuse delay is set to ZERO) and you don't have to relabel.

The medias are deleted from a storage pool by a process of expiration, reclamation, move data and others. As this processes occur and the medias are returned to the scratch pool and because TSM doesn't know it's dealing with a vtl the space in the vtl isn't cleared, the volume will occupy the space until it's reused (as a real tape isn't erased until TSM reuses it). There is an option for some types of vtls (relabelscratch) that can make this an automatic process but not for the emulation I'm using.
 
PREDATAR Control23

I haven't worked with Fujitsu systems, but is there no option to change library to SCSI instead of ACLS emulation?
 
PREDATAR Control23

There is a SCSI emulation but up until the last version didn't supported control path failover and it's limited regarding the number of virtual medias per virtual robotics.
 
Top