Disk sizing, VMS 7.1-2

[Adrian Graham_1]

Folks,

We're upgrading a customer's Alpha 2100 to a DS10, and their disk subsystem from a RA7000 to dual HSZ80 MA2200 controller shelf with 5314 disk shelves.

Given that their SLA with their clients says the OS will never be touched I'm being a tad cheeky by upgrading from their current V7.1 to V7.1-2 though this is necessary for the DS10.

I know the HSZ80 is happy with multi-terabyte LUN sizes as we have others out there full of 146gb drives, but is there a maximum volume size for that version of VMS? Given the availability of brand new drives we're probably stuck with mirror pairs of 72gb each.

Thanks!

PS remember we cannot upgrade VMS...

[Hein van den Heuvel]

For OpenVMS V6.0 and up, you are good with any slice up to 1TB.

FAQ:

http://h71000.www7.hp.com/faq/vmsfaq_013.html#file5

Good luck,
Hein

[Adrian Graham_1]

Brilliant, thanks for that :D

[Bob Blunt]

Adrian, I'd focus on finding the release notes for the version of ACS that you're fielding on the HSZ80s. That should give you a better idea about what disks are supported and what combinations you can build. You can also optimize the way the storage sets are spread across the available SCSI ports to help with performance.

There will probably be a support mismatch between your O/S version and the more recent HSZ program code but you're already hanging over the edge on the O/S version you're stuck with. Be sure, of course, to find all the latest patches for OpenVMS you can for that release. I'd collect all of those available from the FTP site before Sept 18. and install them as you build up the system.

bob

[Art Wiens]

Just curious what type of application / circumstance forces you to stay at 7.1? I'm not judging, I still have VAX v6.2 systems (with a v5.5-2H4 compile environment) and we even have the source code for the one app! I did manage to move our oldest holdout from their 2 node, Alpha 1200 cluster running v7.1-1H1 to our ES47 v8.3 clusters within the last two weeks! Only took 7.5 years! Glacial indeed ... good luck.

Cheers,
Art

[Adrian Graham_1]

Bob: ACS is 8.3Z which is fairly old, the latest I think was 8.6 or 8.8 but yes I need to check that 8.3 will support 72gb drives.

Art: It's purely a written agreement that the version of Oracle they're running is qualified on that version of VMS and neither shall be upgraded because nobody's got the time to do the detailed verification required. It took me some negotiation for the purely -2 hardware release to support the DS10!

A

[Ian Miller.]

Ensure you charge the customer a proper amount reflecting the cost of standing still for 13 years. Perhaps it will encourage them to move into this century.

[Adrian Graham_1]

Hi Ian,

They need to stand still - some of the trials and testing they do last for 10+ years so they can't risk a hardware related foobar in that time...

[Hoff]

You undoubtedly already know this, and know what delusions and illusions are in place.

I'd probably have a long and careful chat with the HP support folks or whomever is going to be supporting these fossils, then. Access to the patches is being restricted, this release is far enough back that there are no new patches, and the six or eight year old gear that was the "upgrade" here is going to have increasing maintenance issues over a decade.

And that you're even asking these questions here in ITRC implies that this project is going to be far more costly, too; that there's insufficient budget for this project up front, which usually then means higher costs over the lifetime of the deployment.

[John McL]

Adrian, give some thought to what might happen with a hardware failure in ...let's say 8 years time. Do you think you'll be able to replace those parts? Will disks of compatible size still be aavailable? Will the network or other comms gear still be available?

I'd be trying to use the latest available hardware and software, and accept the fact that while they'll just about be museum pieces after 10 years, at least you have a reasonable chance of replacing them if something goes wrong.