How do I configure data inputs for .csv files with dynamic field headers for a new event on each line

[fox]

Running 4.2.1, we are monitoring many csv files that differ on listed fields. We have splunk configured to dynamically read the header row for field names. (props.conf: CHECK_FOR_HEADER=TRUE) and this works brilliantly! However we are not seeing the events split correctly - splunk is indexing 256 rows to one event. This is a .csv file with a clear event new line separation...

Has anyone else done this successfully?

Any ideas?

[gkanapathy]

Most likely, Splunk is not detecting a timestamp in your rows. The default rule for Splunk is to merge lines together (SHOULD_LINEMERGE = true), but to split them whenever it detects a date (BREAK_ONLY_BEFORE_DATE = true). The easiest and best way to break on newlines is to simply set SHOULD_LINEMERGE = false, but if there are dates in your data and Splunk isn't finding them, you should also set TIME_FORMAT and TIME_PREFIX and maybe MAX_TIMESTAMP_LOOKAHEAD.

[Ayn]

By default Splunk will merge lines in incoming logs and then break them up according to certain rules. This behavior is controlled by the SHOULD_LINEMERGE directive in props.conf (default is true). Setting SHOULD_LINEMERGE to false will tell Splunk not to combine several lines into a single event, which will give you the behavior you want.