EDIT: I noticed that my REAL record schema was occasionally using integers as the name, and that violated the field name requirements for avro. So, the schema worked once I added a character before the integers. With NiFi's ConvertCSVToAvro, I have not found much guidance or example regarding the Record Schema Property. The documentation says to use an Avro schema, and it seems like a canonical Avro schema does not work. How do I set it up?? My efforts resulted in the caution symbol saying the message below, and the inability to actually use the processor. 'Record schema' validated against '<text of record schema>' Here are the properties as I set them... Record schema: { "type": "record", "name": "sensor_data", "fields": [ { "name": "TimeStamp", "type": "string"}, { "name": "Sensor", "type": "string"}, { "name": "Measure", "type": "string"}, { "name": "MeasureValue", "type": ["int","null"], "default":"null"}, { "name": "AdditionalInfo", "type": ["string","null"], "default":"null"} ] } CSV charset: utf8 CSV delimiter: | CSV quote character: " CSV escape character: \ Use CSV header line: false Lines to skip: 0 Here's some sample data: 2016-02-01T00:03:09Z|Temp_1|TempF|212|"WARNING|Overheating" 2016-02-01T00:05:03Z|Power|N/A||"STATE_CHANGE|Active" 2016-02-02T12:30:00Z|EmployeeCount|0| ... View more