@Asking question with further nuance on topic I believe asked before:
We are deploying in cloud with easy scale-up of instances (and in fact auto-scaling available). We want to use TCPPING for workability/reliability. So obviously we want a configuration file that is instance-agnostic for the cloud, i.e,, with a dynamically modifiable initial_hosts setting. This may be too much to ask for but if not we hvae been led to believe that in jgroups-tcp.xml:
<TCPPING timeout="3000"
initial_hosts="${jgroups.tcpping.initial_hosts:<remote_server_ip1>[7800],<remote_server_ip2>[7800],<remote_server_ip3>[7800]}"
...
that for each instance's copy of jgroups-tcp.xml, remote_server_ip1 above should = localhost for each instance in the cluster (for efficiency of discovery?), i.e., with the balance being the other members of the set, i.e., initial_hosts is different on each instance?
Is there any truth to this for cloud deployment (btw, all our instances are in the same subnet) Or can initial_hosts be setup with the same (albeit fixed) string list of members?
What is best practice for dynamically modifying (automatically not manually) a TCP cluster? Is this even possible?
@
