Hi, I am evaluating a log system solution and then I found Flume. I have a question and have searched on web but I didn't find any discussion about it.
My question is why Flume use an algorithm as "more than half" on fault-tolerant in distributed mode. I think at almost time it is feasible that there is only one server living (or two, dependent on the tolerance) but not a rate of 50%. I know I could fix this problem by myself since I want it, but I want to know whether there are some real reasons that we have to use a rate over 50% instead of a fixed count of living servers? Best Regards xzer
