Hi all, I have a question about load-balancing. I have easily built a cluster with two nodes, but I am wondering how my client should connect to this cluster.
- Run queries against one node (but all data will transit through this node and this way creates a SPOF) - Run queries against an external load balancer (IPVS for example, but load balancer should be known if a node is down) - Client will do the job (like memcached) .. I am using PHP and high level client (pandra) is only in 0.1 version... Can thrift handles this point, ie can I give it a set of nodes to connect to ? Many thanks