Given the following data sample of 4 events where each event has a number immediately after the timestamp that indicates the number of hosts to be listed:
10/07/2013:09:00:00 3 "host1" "host2" "host3" foo
10/07/2013:09:01:00 2 "host1" "host2" bar
10/07/2013:09:02:00 4 "host1" "host2" "host3" "host4" baz
10/07/2013:09:03:00 1 "host1" foobarbaz
So in the first event, there are 3 hosts (host1, host2 and host3) and an unrelated field following the 3 host names (foo).
Is it possible to use rex or props/transforms to intelligently capture the number of hosts expected, then use that number to capture the following host names into a multi-value field, then just pick up where we leave off and extract the fields following the host parade? I guess I could write a custom search command to return all the fields sensibly, but am thinking there might be a clever way to do this inline.
Yes. This produces a multivalued host
field from the search bar:
| stats count| eval _raw="10/07/2013:09:02:00 4 host1 host2 host3 host4 foo bar vi is great" | rex "00\s(?<var>\d+)(?<rest>.*)"| rex field=rest max_match=10 "(?<rest>\S+?) " | eval host=mvindex(rest, 0, var-1)
Put the first rex
as an EXTRACT
in props.conf
, the second as a REPORT
with MV_ADD=10
in props
& transforms.conf
and the last eval
in props as EVAL-host
. Make sure you change the first rex
regex so that it's anchored properly in your data.
Edit: Change max_match to a sensible number, too.
d.
Nice! I've learned something as well!
The first part is easy with regex however I can't think of a way to do the second part in splunk they way you would like.
You can do an extraction for each host in "" as the same field. If they are in "".
First part:
rex ":\d+\s(?P
Second part:
rex "\"(?P