Volume Manager for OpenVMS

[dfox_wa]

I have been told that OpenVMS does not a volume manager software available like Veritas software that is availble for differnt favors of Unix. What I looking for is possibly a third party piece of software that will do the same. Looking to take external storage that several LUNs around 35 GB in size and then put them together as one and present it to the host as one disk. This can be done at the storage level but would like to do it at the host level instead. Does anyone know of way on OpenVMS. More of a Unix person.

[Robert Gezelter]

dfox_wa,

OpenVMS can present three physical volumes (or RAID sets) as a effectively a single volume without additional software, it is referred to as a "FILES-11 Volume Set".

Files will be accessed through a single directory structure that encompasses all three volumes (Admittedly, I am keeping this simple, there is far more detail).

On the device level, such a set is still three volumes, although a variety of backup approaches is possible.

An alternative is to use RAID-1 striping to create a very large volume. This will present as a single, albeit VERY LARGE volume. However, the limit on the size of a single FILES-11 volume does not increase in this case.

I have worked with both many times, each has their uses. One should also consider the issues of when/if the "volume" will be expanded before setting things up. This was an issue I discussed in "Migrating OpenVMS Storage Environments without Interruption or Disruption", at the 2007 HP Technology Forum, slides at http://test.rlgsc.com/hptechnologyforum/2007/1512.html

- Bob Gezelter, http://www.rlgsc.com

[dfox_wa]

Robert,

Thanks for the info. You stated

OpenVMS can present three physical volumes (or RAID sets) as a effectively a single volume without additional software, it is referred to as a "FILES-11 Volume Set".

So in in my case I have lets say five 35 GB luns that come from a EMC frame. I create a diskgroup with a mount point of $1$DGA301 for instance. I initially need 105 GB so I use the above option that you mentioned so it looks like I have 105 GB for the mount point. But four months later I want to grow it by another 35 GB. Is that possible that I can grow that on the fly or unmount the mount point grow it and then remount it and it is now at 140 GB in size. And what is the maximum that I can grow using Files 11 volume set?

[Ian Miller.]

OpenVMS is currently limited to a volume sise of 1Tb. The next version will allow 2Tb volumes.

An additional licence is required to enable the striping software if you want to use a large stripeset. No additional licence is needed for volume sets.

Do see the the referenced presentation and do consider you may save money overall by paying for some consultancy.

[Peter Zeiszler]

I actually had an issue with the binding of volumes. The problem was that we had one HUGE file that took more space than the single volume.

If you are going to be using the volumes for one huge file you might want to investigate that.

Is there something stopping you from having the EMC present the disk as a meta-lun of 105gig?

[dfox_wa]

Peter,

No from the EMC point of view we can present metas that are 105 GB. This is not a technical issue but more of a best pratice/business/political issue. The company I work prefers to create meta devices that are around 35 GB in size 4 way metas made up of ~8.5 GB hypers in size. So on the Unix platforms I just add another 35 GB meta to the volume group and grow into new 35 GB over time. So it is a storage on demand philsophy instead of allocating up front what you need for the next three years to the host. Start out small grow into it.

[Hoff]

OpenVMS volume handling is relatively weak as compared with Unix and Linux boxes; there's no logical volume manager (lvm) analog with the base system, and no host-level disk partitioning support.

Before you decide to add a bound-volume set (MOUNT /BIND, et al) to any OpenVMS environment, do ask around about the limitations and restrictions; the down-side of that approach. (I don't usually recommend the approach, given the limitations and given the capacity of "current" disks.)

The usual HP solution for instantiating synthetic or virtual volumes is via an external storage controller; an EVA series FC SAN controller or other such widget, or a PCI-based or external RAID controller.

36 gigabytes is a rounding error in recent- and current-generation storage; I'm installing one and two terabyte spindles. (The older 1 TB spindles are under the OpenVMS ODS-2 and ODS-5 volume structure 1 TiB addressing limits; the 1.5 and 2 TiB spindles aren't.)

OpenVMS does support on-line extension of disk volumes; what is known as dynamic volume expansion (DVE). This capacity increase is instantiated from the disk controller. DVE is often combined with Dissimilar Device Shadowing (host-based RAID-1) to allow multi-site and multi-host storage with on-line capacity increase.

I've found it can be problematic to (try to) map an Linux or Unix approach or solution onto OpenVMS. (Or to map an OpenVMS solution onto Linux or Unix servers, for that matter.) While many of the basics of the technologies are the same, there can be as many differences here as similarities, and in some significant and relevant areas.

Stephen Hoffman
HoffmanLabs LLC

[marsh_1]

hi,

for info see here for spd on openvms software raid product (raid-0,5 and disk partitioning) :-

http://h71000.www7.hp.com/openvms/storage/raidpage.html

[comarow]

One unmentioned solution is host based volume
sets. They anachronistically duplicate data.

One huge advantage of host based volume shadowing is that the disks can be on different controllers, even on different systems in a wide area cluster. You could loose a data site and have your data up to the mintue in a wide area cluster.

You can create shadow sets of multiple LUNS,
and you can add a new larger disk to shadow
and then increase the size. It's very flexible.

There is a problem with EMC disks on VMS,
and that is they don't support Volume Shadowing on the system disk, in fact, they call it obsolete. From an operations perspective that's just a plain dumb answer.

One huge advantage of Volume Shadowing is that you can shadow an extra disk in, say if it's starting to rack up errors, transparently to the user. And, you can add a member to a shadow set, than take one off line and do an upgrade or make other changes. You have your original disk so back out recovery option can't be beat.

Shadowing has come a long way with bit maps
and only shadowing what has changed. Takes a little learning.

Usually, mirroring has to be on the same
controller, lose that controller, your toast.
You can keep copies of your data with
Data Replicator, but it's not a live working version of the disk. It is cheaper than a multi-site cluster, but what is the cost of downtime?

I might add you can create single member shadow sets, and add drives when you need to.
This is very useful and common. You can replace a drive with no down time, transparently to the user.

Have fun!

Bob Comarow