Conference smurf::ase

1924.0. "uniquely identify a cluster" by DECALP::KLAVINS (Ed Klavins, RTR Engineering) Thu Mar 06 1997 06:07

    Let's say there are 2 TruClusters set up on a LAN. How can I uniquely
    identify each cluster? (from a program -- the intention being that when I 
    run the program it will give me the same cluster id when run on nodes
    in one cluster, and another cluster id when run on nodes in another
    cluster)
    For example, on VMS I can use the cluster group
    number, which must be different across clusters. Is there anything
    similar in TruClusters?
    
    1. One option I was thinking of was using the IP address of the
    cluster_cnx alias. But on further thought, I suspect that this is only
    seen on the memory channel interface, and so must not be unique across
    truclusters.
    
    2. Another option: use cluster incarnation time. In my mind, the chance 
    of two clusters being created at the exact same time is almost nil. I
    could add a further (random) seed to make this even smaller.
    
    Any other thoughts?
    
    thanks, ed

    The clu_get_cluster() function returns a clu_name in the clu_cluster_t
    structure. This seems like a nice unique cluster id. But...
    
    How is this value set in the configuration? Is it set using the
    cluster_map_create utility? If so, then what is returned for clu_name
    in a cluster with 2 ASEs configured?
    
    thanks, ed