Unsolved
This post is more than 5 years old
13 Posts
0
32069
July 12th, 2010 08:00
virtual disk copy usefulness
We've got an MD3000i (2 controllers) with disk copy and snapshot and we'd like to use them with some virtual machines on VMWare ESXi or Xenserver (likely it will be ESXi because Xenserver has stability issues with our servers [PE 6950]).
I'd like to understand how to use disk copy and maybe how to integrate it with snapshot.
With disk copy I can recovery from the destruction of the source disk of the copy, but I have to use the same amount of space of the source disks.
If I need to recover only one VM, on a disk with many other VM images, is it correct that I have to restore the entire disk and so all the VMs?
If so, to allow more fine and tunable restoration process, it could be better to create one disk for each VM, but I couldn't any more add vm virtual disks space on such MD3000 disk, so I'd have to waste space making bigger disks than needed at the moment or I'd risk not having any more space to expand VM disks.
I hope that my explanation is clear, but I doubt because the concept could not be fully clear and my English too.
Then I'd like to know how manage snapshot with disk copy, if they require special management together.
And at the end, it is better to have the source and destination disk of virtual disk copy on different array and different controllers?
Thx to all replying people and best regards to everyone,
Andrea
0 events found


JOHNADCO
2 Intern
•
847 Posts
1
July 12th, 2010 09:00
The real usefulness for us is Enclosure protection. You hang an MD1000 on the last enclosure in the chain and you can get working diskcopies to it, so even if the enclosure comppletely fails you full running copies. No recovery needed, you map the host to it, browse the datastore and add the VM back in like a minute.
If you have some really big stores, it makes it so you only would ever have to restore back to the copy. Think about how long it would take to restore a really big store with millions of files in it.
On the wasted space, Svmotion pretty much makes it a non issue. With no Svmotion, it defiently bares much consideration.
On the MD3000i, I suggest more luns period. The controllers quickly can run into iSCSI reservation issues when multiple VMs are put on the same LUN, on this SAN I'd never share really heavy I/O VM's on the same LUN.
afer
13 Posts
0
July 13th, 2010 03:00
Thank you very very much.
You fully clarified my thoughts because it seems to me that your two first considerations are really sound and rock solid.
We don't have an external MD1000, but a remote Qnap to disk backups and we don't have so many files, it is, actually our five servers have less than 500 GB full backup all together, so it doesn't have much meaning to use the disk copy feature on the MD3000i itself where there are already the same data.
I have to deepen Svmotion, but also if we couldn't use it with ESXi, without disk copy into the same enclosure we'd have the extra space to use some of it in bigger luns.
About luns numerosity I though already to use two luns one for primary access from a controller and one from the other one.
Are there performances advantages to use more of them?
Thanks again and regards,
Andrea
JOHNADCO
2 Intern
•
847 Posts
0
July 13th, 2010 10:00
My therory with this level of SAN as the MD3000i. The more VM's you have on one LUN, the easier it is to saturate the controller. I don't think this is near as big an issue with higher end sans. My example would be our super busy exchange server. If I put any other VM on there, I see iSCSI reservation issues. I think it's because of the relatively slow controllers in the san. We run an imaging system which is super busy both on SQL and flat file image files and it has to be on a lun by itself as well.
Each new lun gets a clean slate of several critical resources.
You see these issues on the host side as increased overall disk latency.
Now I have to say, because we are an SMB we actually give out virutally all dedicated luns for each purpose as there is no danger of running out.
We did massive testing before putting these sans into production. Our configuration(s) is not considered best practice. But I can say that it was they resulted in the way best performance for us and our environment.
afer
13 Posts
0
July 15th, 2010 03:00
http://en.community.dell.com/support-forums/storage/f/1216/p/19338726/19721006.aspx#19721006