Splunk Search

How do I pass an event's field value from a subsearch to an eval statement to retrieve another field?

Explorer

How do I pass an event's field value into a subsearch to retrieve another field?

At the moment, I can't use join because the records at the other sourcetype racks up to millions. Due to limitation, the join command will only return a maximum of 50,000 results to perform the join.

I need a direct search, to eval an extra field for each event using a field from the event.

Example of something I am trying,
Placing employeeID, a field from every event of the main search, into subsearch

index=a sourcetype=sta | eval employeeAddress= [index=b sourcetype=stb empID=$employeeID$ | return empAddress]
Tags (2)
1 Solution

SplunkTrust
SplunkTrust

Well if you're trying to get field values out of Search A index=a sourcetype=sta, and you want to use the field values in there to run another search B, and A might run into the millions of rows, then you can't use a subsearch.

I do however think you have your subsearch syntax backwards. The "first" search Splunk runs is always the inner one, and if I'm reading your question right that would be the index=a sourcetype=sta search. Therefore I think your hypothetical subsearch would look like:

index=b sourcetype=stb [ search index=a sourcetype=sta | rename employeeID as empID | table empID] table empID empAddress]

But again, the inner search will get truncated at 50,000 rows as you say so you can't use subsearches, join, append, etc...

However here's the good news:

1) Just get all the events and let stats sort them out.

(index=b sourcetype=stb empAddress=* empID=* ) OR (index=a sourcetype=sta employeeID=*) | eval empID=if(isnotnull(employeeID),employeeId,empID) | stats values(empAddress) by empID

2) If the search above seems to slow (because it gets many events off disk), then just run it once, or maybe once a day/week, to put the employeeID to EmpAddress mapping in a file based lookup.

Then you'll be able to run very efficient lookups to go from ID to address for your searches going forward.

Basic idea is same as #1, except you tack something like | outputlookup employeeAddresses on the end.

further reading about lookups -

1) http://docs.splunk.com/Documentation/Splunk/6.2.2/Search/Useexternalfieldlookups

2) http://docs.splunk.com/Documentation/Splunk/6.2.2/Knowledge/Usefieldlookupstoaddinformationtoyoureve...

View solution in original post

SplunkTrust
SplunkTrust

Well if you're trying to get field values out of Search A index=a sourcetype=sta, and you want to use the field values in there to run another search B, and A might run into the millions of rows, then you can't use a subsearch.

I do however think you have your subsearch syntax backwards. The "first" search Splunk runs is always the inner one, and if I'm reading your question right that would be the index=a sourcetype=sta search. Therefore I think your hypothetical subsearch would look like:

index=b sourcetype=stb [ search index=a sourcetype=sta | rename employeeID as empID | table empID] table empID empAddress]

But again, the inner search will get truncated at 50,000 rows as you say so you can't use subsearches, join, append, etc...

However here's the good news:

1) Just get all the events and let stats sort them out.

(index=b sourcetype=stb empAddress=* empID=* ) OR (index=a sourcetype=sta employeeID=*) | eval empID=if(isnotnull(employeeID),employeeId,empID) | stats values(empAddress) by empID

2) If the search above seems to slow (because it gets many events off disk), then just run it once, or maybe once a day/week, to put the employeeID to EmpAddress mapping in a file based lookup.

Then you'll be able to run very efficient lookups to go from ID to address for your searches going forward.

Basic idea is same as #1, except you tack something like | outputlookup employeeAddresses on the end.

further reading about lookups -

1) http://docs.splunk.com/Documentation/Splunk/6.2.2/Search/Useexternalfieldlookups

2) http://docs.splunk.com/Documentation/Splunk/6.2.2/Knowledge/Usefieldlookupstoaddinformationtoyoureve...

View solution in original post

Motivator

I propose that you use the map command.

index=a sourcetype=sta|stats count by sourcetype |map search="search index=b sourcetype=stb empID=$employeeID$"|table empAddress

Take this as a template:

index=_internal sourcetype=* user=*|stats count by user sourcetype|map search="search index=_audit user=$user$"|table action

Path Finder

Perfect solution! But how could i use the $user$ token in a dashboard? A dashboard thinks it's an input while it is not.

0 Karma

Motivator
$$user$$