It appears FileReference.java does not like 1 item arrays, and thinks they are field references. Just used a string for a path and that seemed to fix it.
I did have to clear out all the queued data in (for me) /var/lib/logstash/queue/ while the process was stopped, then restart it.
Make sure you also have pipeline workers set to 1, or the other processes will interpret rows other than the first as headers.
Odd, since it did seem to fix it.
Note that the error seems to be thrown by FileReference.java (see line 283) where it looks for ambiguous syntax. It took me a very long time to track it down specifically because, as you mention, it makes little sense that it would be the CSV plugin throwing the error, but here we are.
I believe it's an issue of parsing one item arrays as fields.
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.