Snapclone On EVA

[Mme Jonnaert]

We want to use the snapclone feature on a Eva under hp/ux itanium servers.
We have heard about problems with the controllers and their intensive usage when using this option.
have you informations about that?
thanks
tmj

[Uwe Zessin]

Doing the clone puts an additional burden on the controllers and disks. Also, if I recall correctly, you cannot present the clone to the same server under UP-UX. Well, many operating systems cannot cope with 'redundant' meta data on the disks.

In our lab I once had a Windows server stalling the access to a LUN from which a snapshot was taken, but that was due to an old version of Secure Path.

There might be a problem when you are creating multiple snapshots of the same LUN. There is a special way to toggle the snapshot cache to write-through, but it will decrease performance to the parent LUNs. I would not make such a change without a real need. See if you can test your whole configuration under load before you put it into real use.

I'll be doing a BC EVA implementation with HP-UX PA-RISC servers in about 2-4 weeks. While we very likely will not use clones you might be interested in the results anyway?

[Orrin]

Hi Tmj,

The snapclones are faster for smaller LUN Sizes... on the same DG.

A good alternative si to use the clone feature in Vxfs, you can achieve all the same functions without the additional space overhead of the snapclone.

Hope this helps..

Regards,
Orrin.

[Mike Naime]

Snapshot = Space conservative and almost no time to make. We use this!

Snapclone = Full copy that takes time to make. If you are attempting to make a backup copy of a 500GB database, this can take a couple of hours to create.

While the snapclone is available almost immediately for usage, your EVA performance will be degredated until the copy is completed. We DO NOT use this!

[JJ Urich]

I'd like to humbly correct Uwe's statement about not being able to mount snapclone's on the same host... This most certainly can be done in HPUX. I haven't done it with an EVA, but do it routinely with EMC's, XP's, and VA storage systems. HP wrote a command called vgchgid for when EMC for came out with the BCV (business continuity volume) for hpux 5-10 years ago. This command changes the VGID that is saved in the VGRA (volume group reserved area) or at least that where I think it is stored...

To use it just do "vgchgid /dev/rdsk/c0t0d0" then do your vgimport, vgchange, fsck, mount.

Cheers,

JJ

[Uwe Zessin]

Ah. That's good to know. The BC administration guide explicitly states that it is not supported by BC itself. I will keep this in mind when I get back to the system administrator with whom I should implement BC.

I am not familiar with LVM - what is one supposed to do when the VG is made of multiple PVs?