How to break up a RAID 0 logical drive

[Donald Hill]

We have a pair of RX2800 i2  Itaniums running OpenVMS 8.4 that have all four internal disk drives configured as one logical unit as RAID 0 which is also the system disk.  The question is how can we break up the RAID set into individual disks so that we can have the OS on one disk and application data on the others. We will be using Volume Shadowing of disks between two redundant systems to provide us with our required reliability. 

Thanks

Don

[Hoff · Accepted solution]

Obviously, RAID 0 is striping; all your data is scattered across all your disks, and with no data redundancy provided.  If one disk checks out, every fourth clump of data across the aggregate volume checks out with it.

To off-load the data, use BACKUP or COPY or zip "-V" or otherwise, relocate your data somewhere else, then verify that you have all your data, verify it again, find the RAID volume, confirm the RAID volume is your target and not some other disks on the server that have valuable data, nuke the RAID 0 volume, reconfigure the controller, and re-load the data.  You'll need to use BACKUP /IMAGE or perform an OpenVMS reinstallation to recover or to recreate a bootable system disk here, too.

If you want instructions on how to reconfigure the RAID controller, we'll need some indication of which one is in this particular rx2800 box.  If it's the integrated Smart Array p410i controller, then the following two articles should get you into the range:

http://labs.hoffmanlabs.com/node/1782

http://www.eight-cubed.com/blog/archives/001227.html

edit / ps: MSA$UTIL is the host-level interface for this stuff now, if you're not — as the above expect — working at the EFI level; see http://h71000.www7.hp.com/openvms/products/acuxe/ — but given this is the system disk, you may well be at the EFI level.

[David R. Lennon]

Hi, can I ask for further clarification on your statement "We will be using Volume Shadowing of disks between two redundant systems to provide us with our required reliability. "?

I assume these are standalone systems and not part of a VMS cluster. Volume Shadowing is layered on top of VMS and can only operate on disks seen by a single "system image" of VMS (including a VMScluster).

Did you mean replication on the SAN storage sub-system(s) that will copy changes to another set of disks that could be accessed by a (normally offline) VMS backup system?

Or did you just mean that each standalone system will use volume shadowing to keep a shadow set of two local physical disk members to provide for hardware redundancy?

Of course, you could setup a multi-site disaster tolerant cluster that would provide for a high level of "reliability', with the associated costs and a bit of complexity.

[Donald Hill]

Thanks Hoff that is as I expected.  Yes, it is the integrated controller.  I was hoping not to have to go into the EFI as that is a subject I have only delved into once and find it confusing particularly as I have found very little documentation other than things like help menues which I find too curt.

[Donald Hill]

Hi David

These are stand alone systems but will be clustered using internal local physical disks. 

[Hoff]

---

@Donald Hill wrote:

Thanks Hoff that is as I expected.  Yes, it is the integrated controller.  I was hoping not to have to go into the EFI as that is a subject I have only delved into once and find it confusing particularly as I have found very little documentation other than things like help menues which I find too curt.

---

Ayup.   Jim Duff and I both went through this a while back, had a few chats, and we posted versions of the steps and the command sequences and considerations involved — those are the links that were referenced, and Jim posted one of our #vms IRC chats.  Those will get you into the RAID configuration tool.   Once you're into the RAID configuration tool itself, that's all menu-driven.

[Donald Hill]

Yes, that helps with the how to use the commands to do something but I was thinking more of the when and why.  I.e. and explination of what a unit is, is it the whole space of the physical disks combined or is it the logical equivalent of a physical disk device- things like that.  Also what commands you use to contract a logical drive space or the things to look out for if you do.  Much the same complaint of unix man pages if you don't have books to fill in the holes.

Don

[Hoff]

Whether a unit is a logical (synthetic) disk or a physical disk depends on the particular context.  Without knowing what you're reading about or looking at, I don't know how to better answer that. 

For what you're up to, you're destroying the RAID 0 logical disk (unit?), and creating one or more new logical disks (units?), based on your available physical disks (units?).

If you've not already found the Smart Array docs and if the previous replies and sequences haven't clarified things, then consider engaging some assistance — beyond the Smart Arrays, the configuration of and the migration into HBVS and Clustering and related topics can sometimes be fairly complex undertakings. 

As for Unix man pages, try the modern BSD pages.  They're generally quite good.

[Donald Hill]

Thanks Hoff on the reference to the Smart Array Docs - I had not found them.  And thanks very much on the other references on the configuration of the contorller through the UEFI most helpfull

Don