Problems with different version of the operating system in same cluster

[Heinz W Genhart]

We built a new development cluster with 6 nodes and three system disks. One system disk is OpenVMS 7.3-2, one system disk is OpenVMS 8.3 (alpha) and the third system disk is OpenVMS 8.3-1H1 for IA64. The cluster contains a quorum disk, because it’s planned that we remove 4 of this nodes after we finally finished the migration. I cloned the system disk from 3 existing, well running clusters. The quorum disk is new initialized, not cloned.
I started with the 8.3-1H1 System disk for 3 IA64 Systems. Everything worked as expected. Then I continued with the OpenVMS 7.3-2 System disk. There I encountered the first problem. I could not boot the 7.3-2 alpha. The solution for this problem was:
I had to shut down the whole cluster. I booted the DVD of IA64 System. Then I initialized the Quorumdisk with /NOGPT and finally I booted the OpenVMS 7.3-2 System (with adjusted Votes) as first member into the cluster, mounted the quorumdisk and let create the 7.3-2 system the QUORUM.DAT file. After this I could boot the other 4 Systems too.
Then I continued with the 8.3 System disk for alpha. I could boot one of the two systems into the cluster without problems. The last OpenVMS 8.3 system was unbootable and remained unbootable for a long time.
We did really everything what we could do to solve this problem.

- We exchanged the cloned system disk with another clone. No success
- We exchanged the DS25 with another hardware. No success
- We changed the Lan Interfaces on the new box. No success
- We changed the HBA’s of the new box. No success
- We exchanged the HBA’s on the problem box with the HBA’s of a well running box. No success
- We exchanged the Lan cables of the problem box with the cables of a working box. No success
- During this time I tried several times to boot the box, which has now the Lan adapters, the lan cables and the HBA’s of the problem box. I could always boot this box into it’s own root as well as into the root of the problem box. Only the problem box never came up, even what we did.

Then I did a last try. I set a new cluster group and a new cluster password and then I shut down the whole cluster. After this I tried to boot the problem box first. And see, it worked. After this I could boot the whole cluster.

What I don’t know now is: Was it the cluster group and password which I changed, or was it the boot order which solved the problem?

Actually the problem is solved, but there are several open questions.

Does somebody know, what went wrong?
Why a 7.3-2 System is unable to use a quorum disk initialized on a IA64 System with 8.3-1H1 and initialized with /gpt?
Why I could not boot the second 8.3 Alpha as long as I did not change the cluster group and password?
Today is my last working day. I will go on vacation for the next two weeks and thus I will not be able to respond to this thread until April 20.

Thanks a lot and best regards

Geni

[Jon Pinkley]

The following thread is related, and Geni provided a hint about this thread there, but I don't see a reference to the thread titled "Quorum Disk Compatilbility" here.

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1327584

So readers of this thread should read that first

[Volker Halle]

Geni,

whenever you run into such a situation, where a node seems to hang during boot, consider booting with B -fl r,30000 and capturing the console output. Then force a crash, if the node is hung. And save those crashes and the console output for later analysis, while you try 'other things' to work around the problem.

Without this kind of diagnostic data, all you can expect is wild guesses and speculation...

Have a nice vacation ;-)

Volker.

[Heinz W Genhart]

Hi Volker

we always did a debug boot with bootflags x,30001
I have some of the output logs and we sent some of them already to HP. We are working very close with one of our Swiss OpenVMS Embassadors. So I guess we are very close to HP. But all involved very expirienced people could only speculate until now. I attached a file. This is the only output file I have now at home and I can't get the another outputfiles at customers site, because I will start to my hollydays within the next few hours.
I marked my comments with a line of #####, so you can find it easy.

Regards

Geni

[Peter Zeiszler]

From the init /gpt

If you specify /GPT, the disk might not mount on some systems running older versions of OpenVMS.

[Volker Halle]

Peter,

from OpenVMS V8.2 Release Notes (as referenced in the other thread):

'If you specify /GPT, the disk will not mount on systems running versions older than OpenVMS Version 7.2.'

OpenVMS Alpha V7.3-2 is not 'that old'...

Volker.

[Heinz W Genhart]

we could solve the problem, just 3 hours before my flight schedule. I just forgot to close this thread after my holydays.

Thanks a lot for your help


Regards

Geni