2009-09-17 11:25 PM
Erick Moore posted these comments on my blog article
I'm cross-posting here because I think it is useful feedback.
The checkbox needs to be there BEFORE you create the LUN, and here is why. Let’s say you create a 10GB volume, and now I want to put a 10GB thin provisioned LUN in that volume. From System Manager, you cannot choose the 10GB volume you just created because the LUN will be slightly larger than 10GB.
You may say, well your volume should be larger than your LUN anyway to make room for things like snapshots, but the fact is for something like a LUN that will only contain virtual machines, after a de-dup, the actual used space will be much lower.
While I love what NetApp is doing with this product, it needs work. Here is the process in System Manager for creating a LUN that will be presented to an ESX cluster:
1. Create a 800GB volume
2. Create a 700GB LUN
3. Go back to the LUN and un-check Space Reserved
4. Go back to the volume then resize it to 200GB
5. Exit system manager
6. Go into Filer View or the CLI and map the LUN to your ESX initiator
Step 6 is necessary because System Manager cannot map a LUN that is presented to multiple host on both SAN's because it doesn't properly communicate to its partner when specifying the LUN ID. You will get an error that states "Cluster partner has LUN mapped at this number".
While I am venting the "Create LUN Wizard" is basically useless. I can't even choose the Aggregate to put my volume in! How did THAT get missed? And due to the error listed above the entire process will fail, and nothing will get created. I am anxiously awaiting 2.0 of this product, or even a dot release that corrects some of these annoying issues.
2009-09-17 11:54 PM
This is the reply I put on my blog.
Thanks for the useful feedback.
I'm happy to hear you're enjoying the product direction, but apologize for the frustration.
I've cross-posted your comments to our storage management communities site
I'll try to answer here as well as there, so the info can be more widely shared. Hope that's ok.
When designing the LUN work flow for System Manager, we chose to make the LUNs reserved by default, to ensure novice users didn't run into unexpected out-of-space issues.
As a result, I agree that we made thin LUN provisioning too difficult. We're working on fixing that.
Step 6 you mention is a bug. We're tracking it with NetApp bug ID 362090 which should have a publicly-viewable report soon.
Regarding aggregates, the LUN wizard will only present you with a choice if there is more than one aggregate containing a volume with sufficient free space to hold the reserved LUN.
If there is only one such aggregate, it is chosen implicitly. I understand that this is confusing if you've got more than one aggregate and were hoping for a work flow which is more friendly to thin provisioning.
We're working to address these issues.