topic Multi search / correlate conundrum in Splunk Search

Multi search / correlate conundrum

jeremywebb — Sat, 12 Oct 2019 06:52:16 GMT

Sorry for not spelling the problem out in the title, I'm a bit stuck even for the correct language to describe my puzzle. It's best I explain...

I have one index full of log data like the following:

MessageID User
1 Bill
2 Carol
3 Jane
4 Phil
5 Bill

and also

MessageID Attribute A Attribute B Attribute C
1 Y Y N
2 Y N N
5 Y N N

The Message ID ties together the user and the attributes that I'm interested in as I'd like a table that for each user, counts the number of times they've got a message that has an attribute we are interested in. For example for the above data the table would be:

User Attribute A Attribute B Attribute C
Bill 2 1 0
Carol 0 1 0
Jane 0 0 0
Phil 0 0 0

Or to explain another way...Bill had two messages sent to him, and both had attribute A and one had tribute B. Of Carol's messages, only one had Attribute B.

This is as far as I've got, but it only returns the result for one user, and I'm not sure if JOIN is the right approach for this task!

index=thelog recipient bill | join MessageID [search index=thelog "Attribute A" ] | stats count as Recipient by recipient

Your genius is much appreciated!

Re: Multi search / correlate conundrum

renjith_nair — Sat, 12 Oct 2019 12:41:17 GMT

@jeremywebb,

Try this and test against your data. By looking at your explanation & data, the second row of your result should be Carol 1 0 0.

index=thelog |fields recipient,MessageID,AttributeA,AttributeB,AttributeC
| eventstats values(Attribute*) as Attribute* by MessageID
| stats count(eval(AttributeA="Y")) as AttributeA,count(eval(AttributeB="Y")) as AttributeB,count(eval(AttributeC="Y")) as AttributeC by recipient

Re: Multi search / correlate conundrum

woodcock — Sat, 12 Oct 2019 20:20:34 GMT

Like this:

| makeresults 
| eval _raw="MessageID,User
1,Bill
2,Carol
3,Jane
4,Phil
5,Bill"
| eval sourcetype="A"
| append [| makeresults 
| eval _raw="MessageID,AttributeA,AttributeB,AttributeC
1,Y,Y,N
2,Y,N,N
5,Y,N,N"
| eval sourcetype="B"]
| multikv forceheader=1
| fields - _* sourcetype linecount

| rename COMMENT AS "Everything above generates sample event data; everything below is your solution"

| stats list(*) AS * values(User) AS User BY MessageID
| stats count(eval(AttributeA="Y")) AS AttributeA count(eval(AttributeB="Y")) AS AttributeB count(eval(AttributeC="Y")) AS AttributeC BY User

Re: Multi search / correlate conundrum

jeremywebb — Tue, 22 Oct 2019 07:07:10 GMT

Worked like a charm, and I learnt something from the approach. Thank you.

Re: Multi search / correlate conundrum

jeremywebb — Tue, 22 Oct 2019 07:17:33 GMT

Thanks for this!