I'm using mlcp in load balance, the setup is i have 8 nodes that is load balanced by one ip, mlcp connects to that ip. I kill one node during the ingestion, but mlcp stop and wait for the connection then some documents were not ingested, i did this because i want to perform fail-over test of the marklogic cluster.
How can i obtain 100% without losing data on the process.
You will have much better results with mlcp or other bulk ingestion tools if you do NOT put a load balancer between them and ML. Load balancers are not designed for this type of use case, and mlcp will not function as well (or at all) with a load balencer between it and Marklogic. If you put mlcp on the internal network and give it ANY of the hosts's IP addresses it will by itself discover the other nodes and distribute the upload traffic. WIth a load balender between mlcp and ML it cannot detect or control where each connection is going - at best it gets confused and operates in a non-load balancing fashion. At worse it gets confused and will fail or hang when the load balancer decides to change routing to a new target node.
