Installed 6.0 on AIX in cluster now host server rpt wrong IP

[Tom Canavan]

Our AIX support group installed Cloverleaf 6.0 on a server in a clustered environment. He was logged onto the physical server and not the cluster. Now when I try to access the configuration server via the GUI I’m getting an error reporting the physical IP is refusing the connection. This should be using the clustered IP address and I have a feeling it’s because of the server he was logged into during the install.

Can the IP address be modified or do we need to do the install again from the clustered server?

Topic

[Russ Ross]

See if this gets you over the hump by adding the following to $HCIROOT/server/server.ini

Code:

[firewall]
rmi_exported_server_port=dopivhub

In this example dopivhub is my HA cluster name, also referred to as the service name or address

This is a work around I use in the early stage of my cloverleaf upgrades since HA versions prior to 7 don’t fail-over the service IP transparently.

I also use tricks in my HA scripts to make the cluster name fail-over transparently, but that explaination is another deeper discussion that if you want to have you will need to call me.

I’ve been telling IBM for a decade that the non transparent way IBM/AIX handles the cluster IP on fail-over is not what I consider a legitamte fail-over.

I’ve been told IBM has added my enhancement request in HA 7 to now take care of the cluster IP transparently if you make that selection when setting it up.

I look forward to seeing if all that is true on our futrue uprades, then all this nonsense should be a thing of the past.

I have my doubts a reinstall will change what you are describing, if my suspicioun about the lack of tranparenncy of the cluster IP is getting in your way, which it tends to do with the current deficient design that we’ve had to work around.

FYI – When I do my cloverleaf installs, I do one on each node in my HA cluster to make sure all the system level files that are local to each node are in sync.

Then I just put all the $HCIROOT stuf and below on the shared SAN that moves to the active node.

Russ Ross
RussRoss318@gmail.com

[Peter Heggie]

What is the difference between using the name:

rmi_exported_server_port=dopivhub

verses using the virtual address:

rmi_exported_server_port=xx.xxx.xx.xxx

They are both virtual?

Peter Heggie

[Russ Ross]

I suspect you could use either the virtual cluster name like I did in my example or the Cluster virtual IP.

By the way be sure to stop and start the hostserver after making this change to pick it up.

Russ Ross
RussRoss318@gmail.com

[Tom Canavan]

Thanks for everyone’s help. I was away last week and couldn’t try it until now. I found that using the virtual IP address worked and the GUI displays fine now.

[Author]

Replies