topic Re: Need help with VMS Cluster ... Memory Dump in Operating System - OpenVMS

Need help with VMS Cluster ... Memory Dump

PDP-Fan — Fri, 28 Nov 2008 10:25:21 GMT

Hi there

I have one Alpha ES45 and one ES40 connected through MemoryChannel, directly without a hub. I installed OpenVMS 7.3 on both machines (new installation). Everything works fine until the cluster is activated by Cluster_Config. After the reboot one system starts up and the later one (no matter which one) starts and crashes with a memory dump. This occurrs after joining the cluster.... starting disk serving... error mounting system device.... reset and dump.

I have not defined a quorum disk and there is no shared storage attached. Just the two systems with one disk each and the MC connection.

I think this should work ...?

I have not much experience with Alpha clusters but I set up some VAX clusters some years ago and this was never a problem.

Maybe I have to disable disk serving?

Christian

Re: Need help with VMS Cluster ... Memory Dump

marsh_1 — Fri, 28 Nov 2008 10:44:05 GMT

christian ,

can't remember the commands now at the >>> prompt, but each of the memory channel devices will default to mc0
if memory serves , have you changed one to be mc1 or above ?

Re: Need help with VMS Cluster ... Memory Dump

Volker Halle — Fri, 28 Nov 2008 10:56:28 GMT

Christian,

you are using 2 separate system disks, right ? Did you make the volume labels unique in the cluster or do both system disks have the SAME volume label ? That would explain a PROCGONE crash. early during boot of the 2nd node.

What kind of crash are you seeing ? Just provide the bugcheck message from the console output first.

Volker.

Re: Need help with VMS Cluster ... Memory Dump

labadie_1 — Fri, 28 Nov 2008 11:18:45 GMT

Do you get this message

exit %X7280B4
%MOUNT-F-VOLALRMNT, another volume of same label already mounted

Re: Need help with VMS Cluster ... Memory Dump

Jan van den Ende — Fri, 28 Nov 2008 12:23:43 GMT

Christian,

I think Voler and Gerard (Labadie) hit the right issue.

You wrote
>>>
I installed OpenVMS 7.3 on both machines (new installation).
<<<

In that case, without special extra action, both system disks will have the same label.
Which can not be in a cluster...
After the first node is up, do
$ SET VOLUME SYS$SYSDEVICE/LABEL=
Then REBOOT that node, and boot the cluster.
I would expect that (at least THIS) problem will not show again.

hth

Proost.

Have one on me.

jpe

Re: Need help with VMS Cluster ... Memory Dump

PDP-Fan — Fri, 28 Nov 2008 16:09:19 GMT

I had the same label on both system disks but I already changed one and it did not help. But to be sure, I will reinstall the system with a different label from the beginning.

regarding licenses... I really don't know, do Alpha-VMS cluster licenses exist or are they integrated to the VMS system license?

Re: Need help with VMS Cluster ... Memory Dump

PDP-Fan — Fri, 28 Nov 2008 16:11:13 GMT

The error status is 0072832C

Re: Need help with VMS Cluster ... Memory Dump

Volker Halle — Fri, 28 Nov 2008 16:21:26 GMT

Christian,

%MOUNT-F-DIFVOLMNT, different volume already mounted on this device

Did you use the same disk allocation class (ALLOCLASS) on both systems ? And installed OpenVMS on the same disk (DKAx00) on both Alphas ? Then OpenVMS will complain, that another volume is already mounted on the same device name. Just make sure to use different values of ALLOCLASS on both systems.

On OpenVMS, if something doesn't work, you normally do NOT re-install the system ;-(

Volker.

Re: Need help with VMS Cluster ... Memory Dump

Volker Halle — Fri, 28 Nov 2008 16:24:22 GMT

Christian,

and yes, VMSCLUSTER licenses do exist and need to be installed. But missing licenses do not cause crashes during boot. You will only be able to login on the console without a valid license.

Volker.

Re: Need help with VMS Cluster ... Memory Dump

Hoff — Fri, 28 Nov 2008 16:46:45 GMT

> On OpenVMS, if something doesn't work, you normally do NOT re-install the system ;-(

In general, Volker is correct with this assertion.

That written and unfortunately, this particular clustering environment is one of the few areas where you CAN end up with a need to re-install.

Mess up with the VOTES and EXPECTED_VOTES values or with device allocation classes and such within a cluster environment and particularly within a shared-storage cluster and you can see your disks clobbered. And you CAN end up reinstalling or reloading from your archives.

http://64.223.189.234/node/569
http://64.223.189.234/node/153

I'd strongly encourage a trip through the clustering manuals in the HP documentation set here, as -- if you mess up -- bad things can happen to your data.

Here are the shared files:

http://64.223.189.234/node/169

Re: Need help with VMS Cluster ... Memory Dump

PDP-Fan — Mon, 01 Dec 2008 07:52:55 GMT

I know this is not windows and normally a re-install is not necessary :-) I have a VAX system that runs since 10 years...

But because there is no data on these systems and because it is only a test system, it will be simpler to re-install than to edit all the different values.

I also think that the cause of the problem might be that both system disks have the same name (dka0:) and I already changed the alloclass of one system.
But the problem is still there... that's why I will do a re-install... we will see.

Re: Need help with VMS Cluster ... Memory Dump

Volker Halle — Mon, 01 Dec 2008 07:55:58 GMT

Christian,

how did you 'change to allocation class' ? Eiditing MODPARAMS.DAT and running @SYS$UPDATE:AUTOGEN GETDATA SETPARAMS ?
Does MC SYSGEN SHOW ALLOCLASS show your new allocation class ?

Volker.

Re: Need help with VMS Cluster ... Memory Dump

Jan van den Ende — Mon, 01 Dec 2008 08:29:53 GMT

Christian,

>>>
and I already changed the alloclass of one system.
<<<

Just to make sure: do BOTH systems now have a DIFFERENT, non-ZERO Alloclass? If not, make it so.

Proost.

Have one on me.

jpe

Re: Need help with VMS Cluster ... Memory Dump

PDP-Fan — Mon, 01 Dec 2008 11:06:38 GMT

so far so good... I have set up both systems from zero. The big difference was that I said YES at the question about clustering or not.

Last time I answered no and tried to change that later with all the previously mentioned problems.

guess what... cluster is running now with no further crashes.
I initialized the disks with different labels, alloclass is 0 on both systems but since VMS takes the NODENAME$DEVICE: for the disks, they have different names.

I think I know what could have caused the problems in the first place. Being a bit lazy I made an image backup of the first system disk to start the second system. That might have caused some troubles with identical values that should not be identical :-)

Now I have to search for an OpenVMS cluster license. I have a folder full of old Vaxcluster PAKs but they are useless here.
Since this is a non-productive plaything system... is it still possible to get some evaluation licenses? Where?

Christian

Re: Need help with VMS Cluster ... Memory Dump

Volker Halle — Mon, 01 Dec 2008 11:21:09 GMT

Christian,

if you want to use OpenVMS licenses as a hobbyist, there is the OpenVMS hobbyist program:

http://www.openvmshobbyist.com/news.php

You need to become a member of Connect (previoulsy called DECUS or Encompass) to be able to request OpenVMS hobbyist licenses.

Volker.

Re: Need help with VMS Cluster ... Memory Dump

PDP-Fan — Mon, 01 Dec 2008 16:10:24 GMT

I saw it.... and they want a membership fee :-(

Thanks for your help, I think this problem is solved.