I had an IBM reporting program exporting CSV data with Splunk reading it correctly for a few hours. During this period, I also exported 30 days worth of historical data, and had it all imported correctly.
As soon as the clock ticked past midnight on the 10th of May, everything started going a bit weird. Events from yesterday (10/May/2012) were interpreted as the 12th of May 2010. Events from today are going in as the 12th of May 2011.
Here is an example of CSV data that worked correctly
devicename,volumename,9/05/12 18:05,
And here is an example of CSV data that did not work correctly
devicename,volumename,11/05/12 09:19
As you can see, the date codes in both are in the same format. One worked correctly, one didn't, and I can't work out why.
The data is picked up with this section of inputs.conf
monitor://C:\IBM\TPC\data\log\reports\splunk_mdiskgrp]
alwaysOpenFile= 1
disabled = false
sourcetype=ibm-tpc-mdg
I've tried in vain to add this kind of thing to the props.conf based on what I found online, with no success. Full disclosure: I have no idea what I'm doing.
[ibm-tpc-mdg]
TIME_FORMAT = %d/%m/%y %H:%M
MAX_TIMESTAMP_LOOKAHEAD = 57
TIME_PREFIX = (?i).*?,(?P<FIELDNAME>\d+/\d+/\d+\s+\d+:\d+)(?=,)
I'm hoping someone has seen this kind of thing before!
The TIME_PREFIX
should not have a field-extractig regex in it. It should match the text just BEFORE the timestamp starts, but not the timestamp itself.
Try this:
TIME_PREFIX = ^[^,]+,[^,]+,\s*
Hope this helps,
Kristian
could you post a couple of actual log events?
/k
Did you check this after a restart, and only on events that have been indexed after you made your changes? Because the events that are already in the indexed will not be affected by these changes.
Thanks Kristian, I've given that a shot but it hasn't made any difference.