It seems grok only uses single-precision floating point numbers and their precision is about seven digits. Maybe other parts of Logstash (the mutate filter or the ruby filter) use double-precision floats.
BTW, don't sprinkle your grok expressions with DATA patterns like that. It's a real performance killer. Use a more exact expression or even a csv filter.
About the DATA pattern, I didn't really had a choice since sometimes, a value is so tiny that is it in the form "6.32e-5"...and that is not matched by grok with the BASE16FLOAT pattern...maybe a custom pattern you be better?
Or maybe dissect would be even better? According to the doc, if the lines structure is always the same, dissect could be a good choice.
Is there a way to measure the logstash performances?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.