If i'm running a 5 node zk cluster and a 3 node kafka cluster in dcker on a mesos/marathon environment where my zk and broker nodes are all leveraging local disk on the hosts they are running on is there any value to the local data being preserved across restarts?
In other words.... when a broker node fails and restarts on the same machine does it leverage any of the actual data it has on disk from the prior life or is it assumed to start from new being re-hydrated from the other in sync nodes in the running cluster? Same question for zookeeper... I'm trying to asses the value of local data preservation across marathon re scheduled jobs. Im thinking the data is blown away as much would be stale but not sure since log retentions could be set quite high and having history could make re-sync more efficient by effectively only closing the gap for deltas while down. Anyone know