Decoding wrong values

[stepanurban]

All float values from Kafka printed by logstash are wrong. To Kafka I'm sending following record

ProducerRecord(topic=weather, partition=0, headers=RecordHeaders(headers = [], isReadOnly = true), key=BN, value={"Temperature": 20.49, "Humidity": 56.0, "Pressure": 1011.0, "Wind": 1.0, "Cloudiness": 75.0}, timestamp=1562936770006)

In Kafka it is stored properly but logstash console output is

{
   "@version" => "1",
   "Pressure" => 23.718717575073242,
 "Cloudiness" => -0.09371757507324219,
       "Wind" => 56.0,
 "@timestamp" => 2019-07-12T13:06:10.012Z,
"Temperature" => 0.0,
"Humidity" => -0.09371758997440338
}

My logstash file is

input {
kafka {
codec => avro {
    	schema_uri => "./weather.avsc"
	}
    bootstrap_servers => "localhost:9092"
    topics => ["weather"]
}
}

output {
elasticsearch {
  hosts => ["localhost:9200"]
  index => "weather"
  workers => 1
}
stdout {}
}

and my schema weather.avsc

{
"namespace": "weather",
"type": "record",
"name": "WeatherMessage",
"fields": [
	{
		"name": "Temperature",
		"type": "float"
	},
	{
		"name": "Humidity",
		"type": "float"
	},
	{
		"name": "Pressure",
		"type": "float"
	},
	{
		"name": "Wind",
		"type": "float"
	},
	{
		"name": "Cloudiness",
		"type": "float"
	}
]
}

I'm using logstash 7.2. and io.confluent:kafka-avro-serializer:3.3.3 and 'org.apache.avro', name: 'avro-tools', version: '1.9.0' OS: Ubuntu 18.04.

[colinsurprenant]

@stepanurban thanks for the report. It seems to me that these are probably "normal" float/double representation errors - see https://www.ericlin.me/2016/02/avro-data-types/ for some information on this. If you want to preserve the exact representation I believe you need to use the decimal logical type? To help better diagnose, could you possibly provide an input example that matches the logstash output?