Count of set diff from previous bucket

[jberd126]

I'm trying to produce a report that shows the difference between new and missing IDs from one day to the next day over a period of time. I've looked into this answer but I'm not sure how to apply it to my problem.

The base query

sourcetype=client | bucket _time span=1d | stats values(deviceId) by _time

gives me a set of distinct device IDs. How do I get a set of distinct device IDs from the previous day so that I can get a count of the IDs that are new and removed so that I can get a report such as this?

Date | Total | New | Removed
----------------------------
 x     5       0     1
 x+1   6       2     1
 x+2   9       3     0       

[dmillis · Accepted solution]

This problem is similar to the one described in how-do-i-get-results-in-search-a-that-are-not-in-search-b. Try something like this:

sourcetype=client | eval happened_recently=if(relative_time(now(),"@d")<=_time,1,0) | eval happened_before=if(relative_time(now(),"@d")>_time,1,0) | stats sum(happened_recently) AS recent, sum(happened_before) AS before by deviceId | search recent=0 OR before=0

This will show all the deviceIds which have either shown up in the past, but not today (the "recently missing"), or show up today, but not in the past (the "recently appeared"). You can obviously tweak the final search term to show only "recently missing" or "recently appeared".