This is a "theory" question on how it's done, since I know it doesn't work that way on Teradata Express, but since I'm curious about the details...
In "real" Teradata, what happens when the end-user sees they're going to run out of space and needs to add more?
Let's say we have a few LUNs handed to the AMPs, do they just get grown? Then what? Run a reconfig? Then how/where does the extra space show up?
1. Grow LUN
2. Run reconfig
3. then what?
If DBC's perm space is 500GB, and you add 100GB to the Teradata system, DBC's perm space remains at 500GB, doesn't it? And since it owns all the space, you can't just go in and set DBC's perm space to 600GB, right? Or can you? I guess what I'm getting at is, once more space has been added to the system, how do you make it usable?
Like I said though, this is all just theory. I know on Express, the solution is to install a larger Express VM, and in production, a field engineer might do the work, but I'd still like to know how...
Many thanks in advance!
this is done through TDput (Teradata Parallel Upgrade Tool), it's probably using the RECONFIG utility.
Check the manuals for details.
Btw, usually adding diskspace means adding a new node not simply adding more disks.
Thanks for the response!
So, you're saying that most of the time, if you need more disk space, you add a node? So one doens't usually grow a pdisk LUN or adds a pdisk to an existing AMP? Is it technically possible?
Reconfig looks like it balances the data across new AMPs and such, but the manual for it doesn't say how it handles making the new space available. I assume that in order to make it available, the amount of perm space assigned to DBC would have to increase first, right? The only command I see offhand that might be able to do it would be updatedbc.
It is possible to add disk to an existing configuration if the physical configuration has room. The disks are physically added to the platform and the TVS subsystem is told about their existence. TVS then hands out the space equally to the AMPs within the clique. The same amount of physical disk needs to be added to each clique.
As Dieter said, this is a rare path in the Teradata customer base. In general, when a customer needs more space, they also need more IO and more CPU to process the additional data, users and applications. New node(s) give a balanced complement of all three to the system rahter than expanding disk space without providing more IO and CPU.
In either case, the updatespace utility will change the available space in DBC once the physical upgrade has been completed. Once the space is in DBC it can be handed out to user databases in the usual way.