Toivo,
The data in the queues would be lost if the disk were completely lost. Generally, for a production flow, we would highly recommend using a RAID storage device that provides redundancy. This way, if a disk goes bad the data is still available. The repositories are pluggable, but there are currently only two implementations available: Volatile (in-memory) and File-backed storage. You could also back the repository by a NAS device. I’ve experimented with doing this in EC2, for instance, with an EBS drive, and everything worked well with really good throughput. In this way, if a node were lost, the EBS instance can simply be mapped to a new node, and all is good. Thanks -Mark From: [email protected] Sent: Friday, March 20, 2015 11:27 AM To: [email protected] From Joe mail Dec 17, 2014; 6:14am Re: NiFi architecture „Fault tolerance: If a node dies the data on the node at this time is 'as dead as the node'. “ Dow I understand correctly data in queues (waiting processed by processors) are lost? Are repositories implementations pluggable? Is it possible to implement Flowfile repository and content repository using fault tolerant data store – for example Cassandra? Can this help to avoid losing data in queues? Thanks Toivo -- View this message in context: http://apache-nifi-incubating-developer-list.39713.n7.nabble.com/Fault-tolerance-tp1023.html Sent from the Apache NiFi (incubating) Developer List mailing list archive at Nabble.com.
