How to remove a directory from a File device class ?

Manos

Active Newcomer
Joined
Mar 30, 2017
Messages
7
Reaction score
1
Points
0
PREDATAR Control23

I have a file device class with 6 x 10 TB volumes. As the stgpools on that device class have been reduced all volumes have a 50% usability.
In order to remove one of the devclass volumes I need to move the data to the other volumes.

Any ideas ?
 
PREDATAR Control23

You must replace the entire list of directories with the new one.

So if you have /dir1, /dir2, /dir3, /dir4 currently and want to remove /dir3, you would do:
Code:
update dev {devclassname} directory="/dir1,/dir2,/dir4"

This will not delete any data from /dir3. If there is data still there, you can still read it and access it, new scratches will be created on the directories that remain defined to the devclass.

This means you can do the change now so that no new data is written to /dir3. Then can use MOVE VOLUME to move the data from the /dir3 volumes to existing or scratch volumes in that pool.
 
PREDATAR Control23

You must replace the entire list of directories with the new one.

So if you have /dir1, /dir2, /dir3, /dir4 currently and want to remove /dir3, you would do:
Code:
update dev {devclassname} directory="/dir1,/dir2,/dir4"

This will not delete any data from /dir3. If there is data still there, you can still read it and access it, new scratches will be created on the directories that remain defined to the devclass.

This means you can do the change now so that no new data is written to /dir3. Then can use MOVE VOLUME to move the data from the /dir3 volumes to existing or scratch volumes in that pool.

We do this all the time, and it's completely safe. It can be done while the server is busy or quiet, no difference. I even built a script to automate the process of updating the directory list. The list of directories in the DEVCLASS applies only to writing. Anything previously written remains available until you dispose of it some other way, such as Migration, Reclamation, or Move Data. This is, for instance, a great way to transition to new disks with no disruption.
 
PREDATAR Control23

Just done this today myself and it's more than safe and works a treat.

I do however have a side questions to this.

Using Spectrum Protect 7.1.7
If I have backup data in a file device class using /dir1, /dir2, /dir3, /dir4 and I remove /dir3 & /dir4 from the devclass, move all of the volumes from them into /dir1 & /dir2.

So, If I now add the empty directories /dir3 & /dir4 to a new Container Inline directory storage pool and do a FULL client backup of exactly the same data that's currently occupying the file dev class [/dir1, dir2], will the system work out deduplication efficiently or is my understanding that dedup only works at the individual storage pool level and not across different devclass types?

I ask because I am trying to find the most efficient way of backing up the same data via two different storage sources to one TSM Server. One source will remain dormant (until the retentions expire when eventually it will be blown away) and the other remain active from the initial re-ingestion.?

Thanks
 
PREDATAR Control23

is my understanding that dedup only works at the individual storage pool level
Correct, no dedup across pools.
I ask because I am trying to find the most efficient way of backing up the same data via two different storage sources to one TSM Server. One source will remain dormant (until the retentions expire when eventually it will be blown away) and the other remain active from the initial re-ingestion.?
One large container pool for better deduplication. If some of the data will be kept a really long time, you can look at 8.1.3 or higher for tiering to cloud: https://www.ibm.com/support/knowledgecenter/SSEQVQ_8.1.3/srv.common/r_techchg_tier_813.html
 
Top