MSA1000 on VMS

[Chris Smith_23]

Hi, After several frustrating days struggling with an MSA100 + SAN switch 2/8 on a 2P ES47 I have finally managed to copy a bootable image of patched OpenVMS 7.3-2 onto the mirrored system disk in the MSA1000. However, there is also a 7 disk RAID 5 set in the MSA which VMS can see (with show dev dga) but is unable to init. init $1$DGA200: users gets me an error message '%INIT-F-IVADDR, invalid media address'. HP tech support suggest that I haven't set up a connection for this drive but I believe from every thing I've read, in that atrociously badly written manual WWIDMGR Users' Manual, that I only need to set up a connection for a bootable disk. I can't fathom how to set up two connections to one single port HBA using the CLI command set. Anyone got any ideas? I could attach all my console captures if that would help.

Cheers

Chris Smith

[Uwe Zessin]

Hello Chris,

first thing I would check is whether the Connection profile is set to OpenVMS:

CLI> show connections
Connection Name: xyz_PGA0
___Host WWNN = 20000000-C9382xyz
___Host WWPN = 10000000-C9382xyz
___Profile Name = OpenVMS
___Unit Offset = 0
___Controller 1 Port 1 Status = Online

::one line
CLI> add connection abc_PGB0 WWPN=10000000-C9380abc profile=openvms
::one line
Connection has been added successfully.
Profile openvms is set for the new connection.

CLI>

You're right that you only configure a console device with WWIDMGR for boot and dump purpose, but not for all disks (there isn't even enough space to configure more than 4 LUNs...).

[Chris Smith_23]

Hi Ewi, As you can see from the attachment, I have a valid connection set up. The attachment also shows the output from a 'show dev dga/full'. Both logical drives are visible but the attempt to initialise the 2nd drive fails.

I have been told that the boot device must be dga0: otherwise the 2nd drive will not be visible. Is that the case?

Cheers, Chris

[Uwe Zessin · Accepted solution]

"Ewi"? Whow, you have a strange keyboard mapping or is this line noise ;-)


Is there a chance that the second logical disk is larger than 1 TeraByte
(>= 2^31 blocks)? Unfortunately, that doesn't work on OpenVMS :-(

Otherwise, you really need a certain patch level on OpenVMS so that it works with the latest firmware revision on the MSA1000.

And no, you have been told wrong. I have implemented several OpenVMS systems that boot from devices other than DGA0: (one is a cluster that boots from DGA101:) and they happily detect the remaining devices.

You might have to assign identifiers to the controllers, too - the documentation in this area I have seen so far is conflicting. Some says you need it - some says not needed.

If I recall correctly, the command is:
CLI> set this_controller_id 201
CLI> set other_controller_id 202

[Chris Smith_23]

Sorry Uwe, I was repeating your name to myself as I brought the reply page up and by the time I came to type in my message I only had a phonetic representation of your name in my memory. Symptom of incipient old age!

OK, so I don't have to reconfigure the boot drive to DGA0: Thanks. But, yes the other drive is 7 * 300GB in RAID 5 so approx 6 * 300GB or 1.8TB. Someone else suggested that as the drive had not be written to by the time I issued the init command, the actual initialisation of the RAID set would take quite a long time and that if I was to try it again now it could be completed and I wouldn't see the problem. I can't get along to the site until Thursday morning so I can't test that theory very easily.

Likewise your thoughts on controller ids.

I'll look out the patches. I've already had to apply 4 patches for the fibre HBA
vms732_pcsi-v0100
vms732_update-v0300
vms732_sys-v0700
vms732_fibre_scsi-v0400
(in that order) with rebbots after the 2nd, 3rd & 4th patches.

I will report back any progress after my site visit on Thursday.

Cheers

Chris

[Jan van den Ende]

Chris,

as Uwe expected and you confirmed: the RAID set is simply to big. NO way to get it working!
You will _HAVE_ to split it up into units of < 1 T.

Yes, it is about time for Engeneering to find some clever way to lift another never-expected limit!

Proost.

Have one on me.

jpe

[Uwe Zessin]

No offense taken. I understand that my name can be troublesome to spell for non-germans.

Yes, that's too large for OpenVMS. You can, however, keep the disk array and create _two_ logical disks on those 7 physical disks, each about 900 GigaBytes. Then you could create a multivolume set on OpenVMS. The parity initialization should not affect the ability to write on the logical disk. It is rather the opposite: the MSA1000 is waiting for a first write from the host before it starts the initialization!

[Robert Brooks_1]

Jan wrote . . .

Yes, it is about time for Engeneering to find some clever way to lift another never-expected limit!


It's not likely to happen unless a fair number of customer complain, and state *why* they must have logical volumes larger than 1TB.
It would be a very big project, and it's not clear to product management that it's a worthwhile thing to do -- there are a lot of interesting projects, and don't have the staffing to do everything. As of now, the feeling is that 1TB is sufficient.

If you disagree, please make an official plea to your support centres! Posting notes in public fora like here and comp.os.vms does not help; these places are not viewed with
any regularity by product management.

Your voices on this matter count more than ours!

-- Rob

[Jan van den Ende]

Robert,

thanks for your explanation.

From my (our) perspective, the 1 T is not now a problem, since we do not, and do not intend, to bind disks into big RAID sets.

But, looking at the real world around us with some extrapolation into the future should also be done sometimes by Engeneering.

Look at the figures for the biggest available _SINGLE_ disks in recent years. And place your bet at the date when THAT number will pass 1 T...
By that time, it would be very desirable for the whole VMS community if those devices were not ruled out!

Proost.

Have one ion me.

jpe

[Chris Smith_23]

Hi Uwe & Jan, Thanks for your input. I split the 1.8TB raid into 2 x 0.9TB (approx) sets and VMS is quite happy with that. show dev dg/full gets me the actual number of blocks while show dev dg displays a series of asterisks for the size. Presumably the field specifier for size has overflowed. That doesn't seem to matter as I can now initialise and mount the two drives. I will leave it to the users whether they want the two drives combined into one logical volume. I'm just happy to have the system running to a point where we can load all the users files and accounts.
Thanks again & cheers,
Chris