Manage cluster impersonator

[rieger_1]

Hello!

After the upgrade to openVMS 7.3-2 we have a problem to manage the cluster impersonator.

I read the release notes an found the command to delete a cluster alias address from active system.
ifconfig -alias

I had try these command and was confused,
then no other node work as impersonator.
The cluster alias was died!!

How can I change the imporsonator?
How can I set the interface for waiting to work as impesonator, like "set configuration interface /cluster".

Thanks for information.

[Lokesh_2]

Hi,

Suppose you have two node cluster named NodeA & NodeB. And right now NodeA is cluster impersonator. So, to failover the impersonator to NodeB, you need to give following command on NodeA :

TCPIP SET INTERFACE xxx/NOCLUSTER

(where xxx is your interface name)

This will make NodeB as cluster impersonator. Now if you wish to get back to NodeA as impersonator, then I guess you need to restart TCPIP on NodeA and issue the above command on NodeB.

HTH,
Thanks & regards,
Lokesh Jain

[rieger_1]

Hi,

"TCPIP SET INTERFACE xxx/NOCLUSTER "
was the right command before TCPIP Version 5.4. The release notes wrote "management commands are no longer supported:
* TCPIP SET INTERFACE /NOCLUSTER"

Okay!

Now (TCPIP 5.4) you should use the command ifconfig -alias

But this command don't work in our environment. Nobody will work as TCPIP cluster imersonator. And I don't why.
8-(

I hope you can understand me.

Best regards
Matthias Rieger

[Lokesh_2]

Hi Rieger,

I understand. Though I have never worked on 5.4 , but I can suggest one more thing. Give the following command on the node which you do NOT want to be a cluster impersonator:

TCPIP SET COMMUN/CLUSTER_TIMER=1 .

and following command on node which you want to be a cluster impersonator:

TCPIP SET COMMU/CLUSTER_TIMER=0 .

HTH,
Thanks & regards,
Lokesh Jain

[rieger_1]

Hi Lokesh!

I have problem to check this, because
I had killed our cluster service for one time today.
8-(

The question are:
* How I configure the tcpip services in version 5.4 that the impersonator change automatically(!!!) to next node, if I shut up (in TCPIP) the current impersonator (i.E. for change configuration ore so on)?

* What must I do to configure these node as a standby node for cluster impersonator after the configuarion action (without reboot!!!)?

Best regards
Matthias Rieger

[Garry Fruth]

In the cluster I am testing failsafe in, we are using vms7.3-2 with tcip 5.4. The cluster is heterogeneous; one has ECO4, the other doesn't.

Both nodes are configured using
"ifconfig we1 alias 10.10.10.10/24"

I am able to use "ifconfig we1 -alias 10.10.10.10" and have the ip address fail over to another cluster member. I am using "ifconfig -a" to view the configuration. TCPIP SHOW INTER/FUL shows 10.10.10.10 as the c_addr; but id does not show which is the 'impersonator'. "telnet 10.10.10.10" confirms that the failover worked.

[Andy Bustamante]

Cluster alias seems to no longer function with VMS 7.3-2 and TCPIP 5.4. We migrated to failSAFE IP rather than try to troubleshoot this, which allows a bit more flexabilty.

[Jan van den Ende]

Hi all,

Yes, I agree the Failsafe mechanism IS a step forward, bit I still hate the entire concept of "just" a minor update changing things in such way, that what was the ONLY such functionality before the "minor" change no longer works, while the repacement, that was NOT available before, AND IS ONLY MARGINALLY DOCUMENTED, MUST be configured NOW.

Remeber, the old (and current) VMS adagium:
NO change should break existing code!

Well, to me this change clearly defines TCP/IP to be _NO_ acceptable VMS functionality.
(and it shares this with quite a lot of stuff we "get" from NIX.

Sad, but I regrettably have no other conclusion.

(I think I remarked this before, but in Dutch "U" means "you", and "nix" means "nothing")

Still hoping that those things may one day get better,

Prost.

Have one on me.

jpe