Hello, Each node of the cluster is running a copy of the dataflow that you see in the UI, and the data that is being operated on is what is divided across the nodes (depending on how the data enters the flow).
If a node goes down, the other nodes continue processing, and the data that is on the node that went down is stuck there until the node is brought back up. If the repositories are on mountable drives then you could spin up another node and re-attach those volumes to pick up where the node was, in addition you can use standard RAID configurations to avoid the most common cause of disk failures. There is also a new effort in the community called "stateless nifi" (formerly nifi-fn), which would be an alternative approach and would be able to handle failures differently when the source of data offers some type of replayable acknowledgment protocol (i.e. something like kafka). Thanks, Bryan On Tue, May 14, 2019 at 11:17 PM Jianan Zhang <[email protected]> wrote: > > Hi all, > > We have a nifi cluster with 3 node. If we lost a node, what will happen to > the job running on the disconnected node. > The cluster will do nothing till the node is recovered, or cluster will > reallocate the job running on the dead node to other healthy node. > > Thanks > Jianan
