NiFi stores the info within the repository because it is traversing through the system. There square measure three key repositories:
The flow file repository.
The content repository.
The rootage repository.
As a processor writes information to a flow file, that’s streamed on to the content repository, bearing in mind the processor finishes, it commits the session. This triggers the rootage repository to be updated to incorporate the activities that occurred for that processor and later on, the flow file repository is updated to save lots of track of wherever within the flow the file is. Finally, the flow files are often affected by the likewise as-door-door queue within the flow.
This exaggeration, if NiFi goes the length of at any narrowing, it’ll be adept to resume wherever it left off. This, however, glosses on the extremity of 1 detail, that is that by default following we have a tendency to update the repositories, we have a tendency to write the into to repository however this is {often|This can be} often cached by the OS. Within the row of any failure, this cached information can be speculative if the OS fails on a NiFi. If we have a tendency to set sights on reality nonentity to avoid this caching we are able to set up the repositories within the knife properties file to perpetually adjust to disk. This, however, is often a major hindrance to be in. If lonesome NiFi will the length of this not be problematic in any exaggeration to information, as OS can nonetheless be in command of flushing that cached information to the disk.
