HACMP 5.1/AIX 5.1 ML3. HACMP node re-integration issue.

[ggitlin]

Hello HACMP gurus...,
I am running AIX 5.1 and HACMP 5.1. There are two nodes in the cluster each with it's own resource group. The resource groups are in "cascading without fallback" configuration ("CWOF enabled" parameter in the resource group definition is set to TRUE). Any node can have any or both resource groups.
The problem I am having is when both resource groups are on the same node (one of them failed over to the "backup" node)and the node with the higher resource priority reintegrates into the cluster, it takes back it's own resource group. I thought the CWOF should prevent this from happening?

 

[kardiac]

Hi,

It's maybe stupid but when you restart hacmp on your failed node do you specify you don't want reacquire resources ?
In smitty clstart, field "Reacquire resources after forced down ?" must be to "false"

Bye

 

[ggitlin]

Doesn't this option refer to the state of resource group after they were shut down with the "forced" option (there are three - graceful, takeover, forced)?

 

[kardiac]

True but i have seen some things in redbook "HACMP 5.1 -Administration and troubleshooting guide" as that :

"Set this flag to true if you want HACMP to reacquire
the resources during a startup process. This could
happen if, while the cluster services were stopped
with a forced down option, you have manually
changed any of the cluster resources such as adapters,
disks or applications.
Note: Depending on cluster configuration and state,
the resources may end up on a higher priority node.
Therefore, do not set this option to true if you know
that the application is already running on the higher
priority node."

For me, this option force node to reacquire ressource and that same if ressources are running on a other node.
However I can have wrong. I will test that today if i have time.

 

[ggitlin]

Thank You.

I believe that this option is set to FALSE, though.

 

[kardiac]

Can you send output from this command :
/usr/es/sbin/cluster/utilities/clshowres -g 'ressource group name'