Unsolved
This post is more than 5 years old
5 Posts
0
547
January 14th, 2013 01:00
Converting standard device to virtual provisioning data device
on our VMAX we had 4 disk groups,
each disk groups consists of 33GB hypers on 300GB FC drives.
We had about 130TB of usable space, out of 130TB only half of them
are in use. That means 50% of hypers are unmapped, I'd like to
convert those unmapped hypers to data device in virtual provisioning.
I was thinking of convert standard R5 device to Data device.
Then create Thin pool and bind the Thin dev to the Thin pool.
What is the process of converting standard devices to virtaul provisioning data devices
Is it matter of deleting existing unmap hypers and creating them as 33GB data devices?
8 per drive
Is there any carvet we need to be aware of? In terms of performance and capacity planning?
AranH1
2.2K Posts
0
January 14th, 2013 13:00
You cannot convert a standard device to a data device, you will have to delete the existing devices and then create new Thin DATA Devices (TDATs).
There are performance considerations that apply when planning out DATA devices for the disks. Ideally you should have no more than 8 splits per disk per pool and the TDATs should be as large as possible, and all the TDATs in the Thin Pool must be of the same disk and RAID type, and should be the same size.
The challenge with your existing environment is freeing up enough contiguous space to plan out the TDATs properly. You will want to remove a contiguous range of devices on the disks such that you can create the TDATs following best practices, but this can be very difficult to execute on an exisiting disk provisioned Disk Group (I know, I have done a few of these).
Another approach would be to migrate volumes that are in use to other disk groups so that you can wipe out all the devices and fully configure a Disk Group as TDATs. Then you can migrate volumes into the new Thin Pool from other Disk Groups and peform the same process until you have rebuilt all of your Disk Groups for Thin Pools.
swanliao
5 Posts
1
January 14th, 2013 22:00
Aran Thanks for your reply.
Is it ok to delete the existing devices from SMC or using Symmcli? what would be the easiest way to use SYMCLI command to query list of unmapped devices in each of Disk groups? What command line can I use?
When you said the challenge I will face with existing environment, I am kind of sense that, but if I can delete and re-create the TDEV and understand the raid type and size of TDEV we need, wouldn't it be a simple tasks? Then only concern I have is what happen if I can't free up enough hypers and disks in order to convert to TDEV?
You mentioned another approach would be to migrate volumes that are in use to other disk groups so that I can wipe out all the devices, is this using virtual lun migration (Symmigrate)? is this supported under SMC?
We thinking about free up first stage of disks enough to construct Thin pools and migrate Thick vol to Thin and once we finish migrate the first lot then reclaim those disks and add more disks to the Thin pool.
Is Thin pool suppport non-disruptively add new disks to existing pool?
Thanks
Zikas
278 Posts
0
January 15th, 2013 01:00
Regarding your question, yes you can add new physical disks on a disk group and it is non-disruptive procedure.
Start the Write Balancing manually the night in order to avoid any delays during the working hours.
Also adding nre TDATs on a Thin Pool is a non-disruptive process as well.
Zikas
278 Posts
1
January 15th, 2013 01:00
Hello Sw Li,
i will try to contribe to your answer as well.
The design side has been covered by AranH.
No, in my opinion if you have 4 disks groups, try to tidy up the disks groups, meaning find what standards devices resides on each disk group.
Then try to migrate the standard devices on two disk groups, if it is possible, in order to free up enough space on the other two disk groups.
Of course that means that you have to collect WLA or STP files to find out the workload of the devices and if it is wise to move them on two disk groups.
If, from workload point of view, you can move standard devices on two disks groups then use Symmetrix Optimizer to do the migration.
After you have move all the devices on the two disks groups, then delete all the remaining free space on the other two disk groups that you are going to use for Thin Pools.
In my opinion if you have ESX hosts try to create TDATs devices as RAID5.
In my opinion if you have UNIX/Linux hosts try to create TDATs devices as RAID1.
Then you have to create FAST VP policy for the Thin Pools.
All the above are from my point of view and in my humble opinion.
Best Regards
John