eva snapclone

[ilayy]

hi expert
i need your help with snapclone. what i understood from my reading is that
the blocks that were only mapped to
the virtual disk(source disk) are now mapped to two places - the virtual disk and the snapshot. The mapping is now shared.
when changes are made to a file residing on the virtual disk(source disk) and Before those changes are made to the original disk, the data is moved to the snapshot—a
copy-out. the snapshots mapping has been updated accordingly so that it can find
the data at its new home. so the snapshot record only the changes made to the source disk not a full copy of the source disk so if the source disk fail i have only the changes. is that true?

[Víctor Cespón · Accepted solution]

You're talking about snapshots, not snapclones.

When you create the snapshot, the controllers make a copy of the mapping table (pointers to 2 MB blocks of data). From then on, when a write is sent to the vdisk, the block you're going to overwrite is copied. The pointer on the vdisk points to the new block, the pointer on the snapshot points to the old block.

For the blocks that have not changed, both pointers point to the same place. If something really bad happens to the data on the vdisk, you can select restore and the pointers for the blocks that have been modyfied will be modified to point to the old blocks.

A snapclone is similar, but all the blocks are copied regardless of having being modyfied or not.

[Peter Mattei]

The EVA supports 3 types of Business Copies:

- Snapshot
Create a copy of the mapping table an only write delta of the two vdisks. Method is Copy on Write: the original block first gets copied and then overwritten.

- Snapclone
Starts like a Snapshot but copies all blocks in the background. When finished both vdisks get absolutely independent.

- Mirrorclone
Starts like Snapshot, copies all blocks in the background but vdisks stay paired. When you fracture vdisks the EVA keeps delta block maps for both source and mirrorclone vdisk. You can resynchronize the vdisks at any time in either direction.

Now, regarding data loss - if you loose the source vdisk of a snapshot you will also loose the snapshot since the delta data alone is useless! (It is blocks not files - the EVA has no way of recognizing file systems or structures)

Cheers
Pete

[ilayy]

yes sorry i am talking about snaphot what i didn't understand is why to create multiple snapshot for same VDISK for example
if i create
snapshot1 on friday at 10am and i changed the data at 1 pm so in snapshot1 i have the data of 10 am
than i created snapshot2 and i changed the data at 4pm so in snapshot2 i have the data of 1 pm but also in snapshot1 i have the data of 1 pm so the 2 snapshot contains the same data. please clarify for me why to create a snapshot everyday as as i think all will contain the same data. am i wrong?

[Peter Mattei]

No, this is not correct!

The snapshots will be frozen at the time they are taken (unless you mount them somewhere and update them, which is fully supported).

Assume you just take snapshots and leave them alone.

original vdisk
snapshot 1 taken at 1PM
snapshot 2 taken at 5PM
snapshot 3 taken at 8PM

you will have 4 vdisks
the original which is used by your application and 3 snapshots that have the data of the time they were taken.

You can use the snapshots for any purpose - also for recovery of your original vdisk

Cheers
Pete

[ilayy]

thanks for you answer
you sad that if i loose the source vdisk snapshots are useless (unlike snapclone and mirrorclone) so for exemple if i have raid 5 on the source vdsik and i loose 2 disks in the same RSS which means i lost the raid i cannot recover using snapshot?
2)i have another question if change the place of disks in the enclosure or if i put the disks in a new enclosure the data is lost?

[Víctor Cespón]

Even if you create a snapclone or mirrorclone, if it's on the same disk group as the original vdisk, it will fail too.

To avoid this:

- Create snapclone on another disk group.
or
- Create snapclone in RAID1 (less likely to fail)

But, the risk of double disk failure in a single RSS is very low.

You can change the position of the disks, or move them to another enclosure. The EVA uses the WWID to recognize the disks, does no care about the physical positions.

[ilayy]

sorry for asking more but why if the mirroclone or the snapclone fail if they are in the same disk group as the original vdisk although after they are create they become like a seperated vdisk?and cannot be used for recovering from the failure

[Peter Mattei]

Because all your data within a diskgroup is striped across all disk of that group.
In the unlikely event of a double disk failure the whole disk group gets unabailable: thus all vdisks, snapshots, clones etc.
There are other additional mechanisms to protect disk groups but this would lead too far!

For more in depth info please read the EVA best practice whitepapers on http://h20195.www2.hp.com/V2/erl.aspx?cc=us&lc=en

Search for whitepapers and EVA

Pete

[ilayy]

thanks alot guys for you help