topic Splunk is indexing some of my json objects together in one event in Getting Data In

Splunk is indexing some of my json objects together in one event

lanceblais — Fri, 20 Feb 2015 18:19:09 GMT

I've seen related questions on this subject, but I'm a total newb to splunk so I can't figure out if the problem they're having is the same or not. To make it worse, I have no idea where to put their answers in the config tree!

The semi-problem:

We're dumping JSON into log files on our servers. Everything is properly escaped and 1 JSON object appears per line in the logs.
However, we noticed that Splunk has less records that our other logging stack.

In searching, we realized that 80% or so of our events has a linecount of 1, while others have 2-10 on average and one has 2000.

I'm told I should create a new sourcetype for my log files with a custom LINE_BREAK so that Splunk correctly parses our 1 JSON object per Event.

I'd like to know the following:

1) What could be causing Splunk to not be able to parse my JSON logs, when others (td-agent/logstash) are able to no problem?

2) Is there a key/value format (lots of text like stacktraces) which Splunk could do better with?

3) If JSON is fine, what do I need to modify so that Splunk is able to parse my logs correctly?

Thank you and apologies for the potential dups!

Re: Splunk is indexing some of my json objects together in one event

phoffman_splunk — Fri, 20 Feb 2015 18:31:12 GMT

what is your input stanza for this; and are you defining any props/transforms for this data?

is this coming from a forwarder or straight into a single splunk server/indexer/search head ?

Re: Splunk is indexing some of my json objects together in one event

lanceblais — Fri, 20 Feb 2015 18:54:51 GMT

I haven't touched any config files.

I setup the splunk forwarder for Ubuntu (this is a cloud install) and ran splunk add monitor /path/to/file

If you want the output of some files, let me know.

Re: Splunk is indexing some of my json objects together in one event

somesoni2 — Fri, 20 Feb 2015 20:06:14 GMT

Ans 1: Splunk's built-in sourcetype's have a definite attributes defined and if some of your events don't satisfy all the attributes, you can see random behavior like that. That's why for some non-standard logs, its recommended to create own sourcetype definitions to control the event parsing.

Ans 2: Event parsing doesn't depend of whether its a key-value pair format of JSON. Key value pair format is only good for field extractions (not sure I understood your question, so ignore if it's irrelevant)

Ans 3: You would have to explicitly configure event parsing (and preferably timestamp recongnition) in props.conf on Indexer/ Heavy Forwarder
e.g. for JSON in following format

{ "payload":{ "field1":"Value1", "field2":"Value2", "timestamp":"2015-02-18T13:38:57-0500" }, "event":[ { "field1":"Value1", "field2":"Value2",  "eventserial":"1EWehjkz", "timestamp":"2015-02-18T13:35:48-0500", "status":1 } ,{ "field1":"Value1", "field2":"Value2",  "eventserial":"1EWehjkz", "timestamp":"2015-02-18T13:35:48-0500", "status":2 } ] }

[YourSoureType]
pulldown_type = true
INDEXED_EXTRACTIONS = JSON
KV_MODE = none
BREAK_ONLY_BEFORE_DATE = false
BREAK_ONLY_BEFORE = {.*"payload"
TIME_PREFIX = \"timestamp\"

Re: Splunk is indexing some of my json objects together in one event

thisissplunk — Mon, 18 Jan 2016 19:02:12 GMT

Did you ever get this figured out? I'm having the same problem.