Is there a TSM Limit on Disk Volumes?

[rvillano]

Hello To All,

I have a kinda unique problem that I need your help with. Here's the issue:

I have a customer with a very large amount of medical data 53TB stored currently on lto1 tapes in a 3584 library. the problem is the recall of this medical data by the user community does not perform to the SLA. the data retrevial takes to long most times as it is often spread over multiple tapes.

As an emergency measure the customer has purchased/rented 60TB of Disk storage on IBM 4000 series and an F20 shark. I want to do a move nodedata from the existing primary tape pool to a primary disk pool with enough volumes defined to move all the data from lto1's.

The AIX 5.3ML5 JFS2 file system does indicate the luns are active and I can write test files and a "DF -g" command displays the correct file space allocation size so... all seems good on the OS side! the TSM server is 5.2.7.3
which I will upgrade to 5.4 as soon as I get this issue stablized.

eventually this data will be moved back to LTO3 tapes via HSM. Now Here's my current problem:

I am using the DISK device class and created a std disk stg pool and began adding volumes via the TSM GUI and was able to create (3) 2TB disk volumes.
The server was rebooted to apply some OS patches for vpaths and from that point on I can no longer create 2TB volumes. I tried using the command line and also tried DSMFMT but with this command I recieved the error that DSMFMT can only format 69,xxx,xxx,xxx bytes max? well I can still create 500GB volumes but at that rate I will have to create 100 500GB volumes to get to 50TB. any ideas as to why I cant create 2TB volumes any more?

I have been concidering blowing this all away so far and using raw volumes but unsure if that will fix the problem. also think of using a file type device class but cant create a file device class larger than 63GB /file there are gonna be a lot of file volumes if that is as large as I can make a file device.

any advice or sugestions will be much appreciated!!!

thanks in advance,
Bob

 

[chad_small]

I was once told by IBM/Tivoli that TSM starts to choke when the volume (NOT THE STORAGE POOL) gets larger than 500GB. Now I would use RAW for speed of creation, but this solution really needs an HSM or archiving tool like Content Manager from IBM. I would also ask if collocation is in use? If not then you have the answer as to why restores take so long. Granted tape use increases but it can be a lot cheaper than disk solutions.

 

[rvillano]

This Storage pool only has One Client backing up to it so in effect it is a colocated storage pool. the problem is the data is still spread over many tapes as these studies get updated by the medical profession. this is PACS data if you or anyone else reads this post. It will be processed using HSM soon but for now it needs to be on Disk volumes instead of LTO1 tape to resolve the immediate problem.

Bob..

 

[schmeckendeugler]

are they "COPY" pools?

I recently am trying to test getting data off tape, and putting into a DISK storage pool; but for some reason still having trouble creating a storage pool of type DISK that is not a "PRIMARY" pool, rather than a "COPY" pool. and it looks like I can only get stuff from tape, to a "COPY" pool...

but I'm a noob so I am probably doing something way wrong.

 

[schmeckendeugler]

why not

Is there any operational reason you want a few large files, vs. many "small" ones? or is it just a personal preference? Or you just want to make sure you don't hit some tivoli limit, like, "A pool may only contain 256 volumes."?

 

[moon-buddy]

rvillano,

You would be going in the right direction by having the data on disk for immediate access. The only issue you would have to answer is "how big the disk size should be" to address the current need.

Long term solution for files that are accessed frequently, and therefore changed frequently, would be HSM as you have mentioned. As Chad had cited, Content Manager would be a good alternate solution.

 

[rvillano]

Thank to all who have provided their input! Ok I have soved the problem but in a different way. the best solution to creating a primary disk pool with 57tB of volumes is to use RAW volumes. the only limitation found for this was 1TB max for the raw volume size so it took 57 aix volume defs and 57 TSM volume defs to complete but... the time to create a raw volume def was almost instant! compared to using a jfs2 file system!!! the other issue was that for some reason while trying to use a JFS2 file system I created 3 2 TB volumes but from that point on the volume creations were limited to smaller and smaller volumes???? why I dont know but thats what I experienced.

Now to the reply about copy pool question in this post... if the storage pool in question is a copy pool then the volumes or node data to move must stay in a copy pool. now it might be possible to create a file device class and create another copy pool for that data there if thats what you are wanting but.. I dont understand why you would want to do that unless you were planning for a dr test because tsm will never ask for a copy pool tape during a restore unless there is a problem with the primary tape pool volumes. I am certainly no expert with TSM so if anyone else has any input please provide help for this question.

Thanks again
Bob