New Cluser Member doesn't see EVA devices correctly

[Karen Lee_3]

I am creating a cluster and adding the first member this weekend. My existing DS25 is connected to my EVA and I want to add a new ES45 member. My EVA administrator has presented all the existing disks to the new ES45 but at the boot prompt they look odd.

First of all - there are 2 entires for each device. I had him present DGA1, and DGA3 to the new server - here is what i see - is this correct?

dga14720.14.0.10.0 $0$DGA14720 COMPAQ HSV110 (C)COMPAQ 30

dga14720.15.0.10.0 $0$DGA14720 COMPAQ HSV110 (C)COMPAQ 3028

dga25763.14.0.10.0 $0$DGA25763 COMPAQ HSV110 (C)COMPAQ 3028

dga25763.15.0.10.0 $0$DGA25763 COMPAQ HSV110 (C)COMPAQ 3028

[Art Wiens]

Two entries per I believe is the representation of the two paths to your EVA/device. As far as the DGA numbers being "unusual", I believe that's the way they appear before running WWIDmgr from the console. And as far as I can recall, it doesn't really matter what they look like at the console prompt, VMS will name them correctly once it comes up.

Art

[Karen Lee_3]

what do i set my bootdef_dev to?

[Art Wiens]

I don't think I'm overstepping any bounds here, the manual I'm posting (WWIDMGR Usersâ Manual) is from an HP site somewhere.

Good bedtime reading if you're having trouble sleeping ;-)

Art

[Mark Garrett]

After you initially do you widdmgr -udid etc do an init before displaying the devices. For some reason they look odd untill you init. See if that make you world look right?
for the bootdef_dev you need to list both paths if you expect it to boot from either, after the boot the OS will use both paths so it will work mostly if you put one, I highly recomend you put both though.

[Volker Halle]

Karen,

could these 'unusual' unit ids be the result of using '- item n' instead of '-udid n' when configuring the disks with WWIDMGR ?

Normally you would do:

>>> wwidmgr -show wwid
>>> wwidmgr -quickset -udid n

(where n is the unit numer of your system disk).

Volker.

[Karen Lee_3]

The disks listed in my example are production EVA devices already seen by the existing cluster member(AVMS).

There are 2 devices presented to both the existing cluster member(AVMS) and the new cluster member(BVMS).

AVMS see "DGA1", and "DGA3" at the >>> prompt; BVMS shows what you see above.
I had the EVA config. tech check the OSID and they both point to "VMS". I have powered off and on several times.

This is really baffling.

[Art Wiens]

I have seen the same thing with my HSG80 while adding a DS10. Read the manual about how to use wwidmgr (it's not difficult at all). Once you go through the few steps required, you will not be baffled anymore, your SRM representation of the DGA devices will be "normal".

Art

[Sheldon Smith]

You only need to use WWIDMGR to add the disk or disks from which you plan to boot. Once OpenVMS is up, it will find any others. And, you are either in Console mode, OR "wwidmgr" mode. You need to INIT the system to be able to "switch". And switching consists of either issuing standard console commands, or issuing "wwidmgr" commands.

Assuming your switch zones and EVA hosts are defined correctly, to be able to boot from DGA1, the sequence would be:

>>> init
... wait ...
>>> wwidmgr -clear all !! clean out any previous settings
>>> wwidmgr -quickset -udid 1 !! set up for Unit ID 1
>>> wwidmgr -show ev !! to see what got set up in wwidmgr
>>> show dev !!! will look very strange here
>>> init
... wait ...
>>> show dev !!! will look normal now
>>> set bootdef_dev dg*1.*.*.*.*
>>> show bootdef_dev
>>> boot dga1

Also, if during your initialization you see "PGan: Nvram read failed", you need to init the adapters ONCE:
>>> wwidmgr -set adapter -item 9999 -topo fabric

[Cass Witkowski]

If you do a wwidmgr -show wwid -udid xxx -full where xxx is the vms identifier of the disk you will see that the disk has two names one is the vms identifier and one is the console name. What you are seeing is the console name.

I have talked to HP about this and the only way we have found to make the SRM console show the vms identifier is to do a wwidmgr -quickset on each disk you need to boot from.

If you have a single system disk then you just need to do the one quickset.

I.e. $1$dga1000 is the system disk

>wwidmgr -quickset -udid 1000

If you have a shadowed systems disk dga1000, and dga1100 then

>wwidmgr -quickset -udid 1000
>wwidmgr -quickset -udid 1100

If you have these disks on different disk subsystems then use the wwidmgr -set -port command to set the correct ports (N* console variables)

Cass