RX2660 and device naming under VMS

[R. Verkerk]

Hello,

Is there a way before an actual VMS installation to know how the devices are named in VMS 8.3 for Itanium.

If I would put in for instance three SCSI controllers how would they be named? IS the bus scanning lineair or is the device order different.

gr,

Robert Verkerk

[Vladimir Fabecic]

Boot from installation CD and choose DCL prompt.
You will see how devices are named ($ SHOW DEVICE).
Not sure if I understood your point of linear, but it should be linear.
But why should device naming be a problem?
You can always define a logical name.

[R. Verkerk]

Hello,

Thanks for the reply. I know I could do that if I have such a system.
But I have to make/adapt some software before I have my hands on the system. That's why I asked.

With Lineair I mean that if you have for instance the RX3600 and you fill slot 1 to 4 with the same dual interface ethernet cards the device names are different than you expect them to be. It will not be EIAa to EIH starting from slot 1. They are mixed up.

gr,

Robert Verkerk

[Jan van den Ende]

Rob,

>>>
But I have to make/adapt some software before I have my hands on the system. That's why I asked.
<<<

That is exactly what logical names are for!

Make your software completely self-consistent within its (set of) Logical Name(s), and have an installation instruction saying they have to define those names to point to whatever is appropriate in the target system.
That way you can deploy the software on any VMS system (of a compliant VMS version)m without having to bother about the specific (future? maybe not even existing yet) hardware config.


Really, consider this _THE MOST IMPORTANT_ instruction for _ANY_ VMS software!

Proost.

Have one on me.

jpe

[R. Verkerk]

Hello,

I have to make for instance a TCPIP config before I have the system.
If I would use logical names than for instance a logical EIA would point to EID and logical EID might point to EIC.

This would confuse everyone using this system.

The same goes for disks and their names:

We mount the disks and assign logicals. But you need to now he devices before you can assign logicals.

Would you like a logical DKA to point to DKD devices? I would not like that!

So that's why I have the question:
Does anybody have the sizing order of the RX2660 in OpenVMS?

gr,

Robert Verkerk

[Hein van den Heuvel]

>> I have to make for instance a TCPIP config before I have the system.

The network interface name should not be used by the applicaiton, only by the configuration process. For those I think most of us would accept the real device name thanks to limited usage, contained in just 1 or 2 files on the system.
Your preparation work would use temporary names, edited later.

>> If I would use logical names than for instance a logical EIA would point to EID and logical EID might point to EIC.

Those would be 'illogical' names. It does happen, but it is way to confusing.
If you are going to call them anything call them 'cluster_network', 'management_network', 'production_network', 'worl_wild_wetwork'


>> We mount the disks and assign logicals. But you need to now he devices before you can assign logicals.

But that's just in one place.
The purpose of logical names is to allow the system to use a pre-chosen name 'everywhere' (in sysuaf, in the application, ...) and to only need to change one line (per device) if the underlying resource for the name changes.

Making the mount command use the physical name satisfies this. Just use a 'temp' name in the preparation work: 'dk_aap:', 'dk_noot:', 'dk_mies:'
During the real installation, look for the dk_ strings in the command file.

>> Would you like a logical DKA to point to DKD devices? I would not like that!

Agreed.
You might temporary call them DK_A and DK_B though.

Groetjes,
Hein.

[Jan van den Ende]

Rob,

For application aapnootmies (hi. Hein!) use aapnootmies_EXE, aapnootmies_DATA, aapnootmies_LOG, aapnootmies_watdanooknogmeer etc.

In your installation instructions you have.

- create :[aapnootmies.EXE] and move the EXEs there

In application boot procedure (called from system boot, either SYS$SYSTARTUP_VMS or a SYSMAN startup phase, whatever the site's preference)

$DEFINE/SYSTEM/EXEC aapnootmies_EXE :[aapnootmies.EXE]
- the same for DATA, LOG, watdanooknogmeer
Note well: a_disk CAN be, but absolutely need NOT be, the same.

- define (or have Network Administration do that) aapnootmies as a DNS record pointing to the system that currently offers application aapnootmies.
- have the users make connection to aapnootmies

Now, whenever you (for whatever unexpected future reason) have to (temporarily?) move the application to anoter machine, at the moment of switching (back?) all that needs to be done is changing the DNS name, and the users will never even notice.
When you (have to) move (part of) the application, move the files, change the logical name, and the users will never notice.

If your installation has to be "dummy run", than you will have to put some extra effort in the installation procedure, and have the installation "ask" for the target device(s), and generate the rest from there.

hth

Proost.

Have one on me.

jpe

[Hoff]

There's no easy nor standard off-line/remote TCP/IP configuration, though there have been a few hack-arounds discussed here. (There was one a few months back, though I don't immediately see it.)

If you can't use DHCP or IPv6 for this host, you're in a somewhat difficult position. A remote console makes this configuration task somewhat easier, but replication and deployment of OpenVMS itself is nowhere near as automated or as easy as it should be.

The closest approach I've seen here is to use the factory-installed software -- or your own version of same -- to bootstrap the deployment environment. The sneaky approach is to pre-set TCP/IP Services to use DHCP, have the client box wake up and start up, then to have the client pull down its configuration from (where-ever) and have it re-register that. But there's unfortunately no easy (supported) way to automate TCPIP$CONFIG.

Unless you're reloading these boxes often and/or you're deploying a whole lot of these boxes, the customization probably won't pay off. But it's an option.

OpenVMS just doesn't do grid-style load-and-go computing all that well.

And as others have stated, logical names are the way to go for disk devices and such.

[R. Verkerk]

Hello,

Thanks for all the replies.

We are going to install two extra SCSI cards in the machine. I wanted to know if they are sized before the internal SAS disks or after and how they are configured/sized.

So are the internal disk's: DKA and external disks DKB, DKC

Or is it totally different.

gr,

Robert Verkerk.

[Carl Bennett]

Rob,
The last couple of Rx3600's that I worked on had the internal SAS cards come up as DKC's.
It's going to be pretty hard to figure out ahead of time on a consitent basis what letter will be assigned to a controller especially when you start throwing cards in for the odd tape drive here and there.
With our application we have created an INI file that lists all of the disk drives and logical names that we intend to use and then call that before starting the main application.
In this way, we can update the application to deal with new controllers, drives, etc without having to rewrite anything.
For example, I might have a volume called STMT_DISK defined in the INI file as DKA0. In a year, when that piece of the application moves to DKB0, I update the INI file, reload all the values to the system logical names table with a simple open /read loop, and the application never has to be aware of the change.