Has anyone out there tried creating a user-defined storage pool
I have read through the Managing Celerra Volumes with AVM and the Managing Celerra Volumes and File Systems Manually (at the 5.5 level) and it appears to me that there is just enough info out there to get oneself in trouble as there are no complete and thorough examples of creating a user-defined pool from A-Z.
I'm guessing that the first step would be to add a few spA/B balanced LUNS to the Celerra
I guess my next question is at what the next step is. On page 32 of the aforementioned AVM manual it indicates that when you define the user-defined storage pool you assign disk volumes (i.e. d126,d127). I'm assuming this is done to create the affinity between those newly added disks and the storage pool you are trying to create. I'm also guessing that you would want to set the -default_slice_flag to yes at that point.
The docs state that you can add disk vols, meta vols, stripe, or slice vols to the pool.
However, on page 7 of the Manual docs it indicates that a metavolume is required to create a file system.
I'm suffering from a disconnect here (caused by poor technical documentation).
Do you have to then create meta vols with the option of slicing and striping, and then also add them to the pool? Or do they have an affinity to the pool if their underlying Dxxx volumes are associated with the pool.
I'm guessing some sort of meta volumes have to be in the pool prior to the creation of the first file system in the pool.
Sounds like your question and my problem (thread - Problem making new NAS POOL) have common roots; lack of good documentation on creating pools. My guesses were similar to yours and I put a pair of LUNs in my Storage Group for Celerra and it created a system defined pool with no members! If meta volumes are the answer, and I suspect they are, are they meta'ed in Navisphere or on the Celerra as they used to be? If so, how do you do that?
Sorry to not have help on your question; I think I'm in the same boat. Harold
Where I think you are working with a system defined pool, the user defined pool has a few more challenges in that AVM is not there to create the slices, stripes, and metas for you when you allocate a file system.
You have to manually specify disks to the user defined pool before they are grabbed by AVM and dropped into the system pools.
Regardless of how you cut the storage in Navisphere, the Celerra has the ability to define dlices, stripes and metas (a set of the aforementioned slices and stripes) on top of the LUNS you present.
It could be that the system defined pools are not populated until they allocate their first file system. In fact I'm hoping that is the case, otherwise I wouldn't be able to cut any R5 disk and put it in a user defined pool of my choosing.
My ultimate goal is to set aside storage to be dedicated for snapsure such that it wouldh't infringe on the system defined pools we use for file extentions. I can manually create a snapshot and specify the pool that I want. Any scheduled snapshots that occur after that will pull storage out of that same pool.
I've yet to get a response on the user-defined pools except your response, so I'm guessing that most folks use the system, and rightfully so since it is easier to use.
I do agree on the poor documentation as it offers only a lick and a promise. Rick A
To create a user defined pool is fairly simple through the GUI. Simple go to pools, and create a new one. You can add disk volumes, slice volumes, stripe volumes, or metavolumes to this pool. Be aware though that the rules AVM uses to create filesystems in the system defined pools do not apply to user defined pools. In the system defined pool for clar_r5_performance for instance, it takes 4 like sized LUNS, creates slices on them, and then a stripe and then a meta to satisfy your storage request. In a user defined pool, it will take all the storage from the first item in the pool that it can use to satisfy the request. To effectively use a user defined pool, you must apply the principles of the AVM profile up to the point of creating the filessytem before adding it to the user defined pool. Once in the pool, you can create/extend your filesystems through the GUI by designating the user defined pool as the source.
User defined pools are a great tool, but an understanding of the storage is needed to optimize how they are used. This information is difficult to find in any one source, and is much to detailed to place in this thread. I would suggest contacting your local team and inquiring on the ability for an EMC representative to consult with you on how to best utilize this feature. It si a very powerful one, and I use it consistently, especially when using iSCSI LUNS for Exchange/SQL. It allows me to isolate specific spindles for logs and data, and insure the correct number of spindles are behind these LUNS to handle the load being generated by the host.
i've seen some documentation that states that AVM can be used with user defined pools ..is that correct ? For example i have a DMX3 with say 146G drives, 300G drives and 500G ATA drives. I want to create a custom pool for each type of drives to provide tiered storage. So can i create these custom pools and populate each one with one specific type of drives, and then let AVM to do all slicing/striping/metaing on its own just like it does for system defined pools ?
Adding the devices to the pools allows you to use the filesystem GUI to create the filessytems, but there is a profile that is used to create the underlying stripe volumes, and metavolumes(Logical Volume manager) that is only enabled on the default system pools. It is not applied to the custom pools.
So...Lets say you add 4 disk volumes of 100GB each to a pool named test. You want to create a 50 GB filesystem. In the Standard pool, this filesystem would be created over all 4 disks by creating a stripe volume of the disk volumes, then creating a slice volume the size you need for the filesystem out of the stripe volume, and then creating the metavolume/filesystem. In your custom pool, it would create the filesystem on a slice of the first disk in the pool only. No stripe volume, etc.
To correct this behaviour, you would have to manually create the stripe volume and then add the stripe volume to the pool instead of the individual disk volumes. This will cause the GUI to create the 50GB slice, and then create the metavolume/filesystem over the 4 drives. Since the 50GB slice is a slice of the stripe, the filesystem will now be spread across the 4 disk volumes.
In other words, to use custom pools, you have to create the underlying volume infrastructure in the configuration you desire the filesystem to be spread across before adding it to the custom pool.
FresnoRed
90 Posts
439
0
Posted January 30th, 2007 07:00
Sorry to not have help on your question; I think I'm in the same boat.
Harold