Is it possible to set up a CA RA such that a load balancer sits in between the connections from agents to execution servers?
i.e. if you have execution server nes1 and execution server nes2, these servers and their nimi ports (nes1:6600, nes2:6600) would be configured as members of a load balancer/VIP of some kind (assume the load balancer/VIP is called nes-lb-1. The agents would then be configured with a supernode of the address of nes-lb-1 in their nimi config.
I'm aware of the HA multiple supernode connection that you can setup in an agent's nimi config, and we are using that. I'm asking about this load balancer setup as it may be useful for us in certain scenarios we might come across in future
I don't think as of now we have this possibility to have LB in between execution servers and agents. If I understand your requirement you are looking for high availability of the agents, but from above example you provided if you still see the context switch to new NES is only happening via Agent (because you are keeping two different values in nimi_config for respective NES's).
From my view point the LB solution best fits, if my Agent just have one entries for supernode and LB take care of it switching based on NES availability, which I haven't came across yet.
The high availability I have seen is to have two pair of NES for each sets of agent which make sure that the agents are always highly available, as they do context switch to different supernode.
I think my misread it, what you are saying "assume the load balancer/VIP is called nes-lb-1. The agents would then be configured with a supernode of the address of nes-lb-1 in their nimi config"
Looks what you are recommending in theory viable but I haven't came across this during my exposure.