I am currently trying to read in data from a .csv that has a timestamp column. When I upload the file and go to preview mode, I'm setting the custom timestamp field to my column. The splunk timestamp reads that data and sets the timestamp correctly for about half the values and then fails to read the data correctly for the other half
Example:
Timestamp EVENT_DATETIME
9/6/14 12:00:00.000 PM 6.9.14 12:00
9/6/14 4:00:00.000 PM 6.9.14 16:00
9/6/14 8:00:00.000 PM 6.9.14 20:00
9/9/01 9:12:20.000 AM 7.9.14 0:00
9/9/01 12:02:00.000 PM 7.9.14 4:00
As you can see, the timestamp is formatted correctly for the first three entries then incorrectly for the last two.
No idea why this is happening...
The default timestamp behavior for Splunk is to guess the times and assume that they generally go forwards, because it was designed for logfile handling. For data that jumps back 13 years, you may have to set a TIME_FORMAT for the data, and you have have to adjust MAX_DAYS_AGO as well as the MAX_DIFF_SECS_AGO / MAX_DIFF_SECS_HENCE. DAYS_AGO puts a bound on the expected difference betwen file modtime and the timestamps, DIFF_SECS puts a bound on how far timestamps are expected to jump from event to event.
Start with a TIME_FORMAT though.