topic Re: How to search the count of adds/removes (new hosts vs host decoms) month on month? in Splunk Search

How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Fri, 10 Jun 2016 13:02:38 GMT

What I want is to many adds/removes (new hosts vs host decoms) month on month

index=* | stats dc(Host_Name) by date_month

So take Jan ... there are 160226 distinct hosts
Feb number is 162359
That could be something like 190 hosts were removed and 2323 were added in Feb

Thanks

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

somesoni2 — Fri, 10 Jun 2016 13:33:04 GMT

Assuming you're just comparing host count for (last) two consecutive month, give this a shot

(Based on your example query, you're not using splunk's default host field, so this will be slower)
Updated

index=* earliest=-2mon@mon latest=@mon | stats values(date_month) as month by host | eval Type=case(mvcount(month)=1 AND month=relative_time(now(),"-2mon@mon"),"Removed",mvcount(month)=1 AND month=relative_time(now(),"@mon"),"Added",1=1,"Same" ) | chart count by Type

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Fri, 10 Jun 2016 13:38:54 GMT

Error in 'eval' command: The expression is malformed. Expected ).

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

Richfez — Fri, 10 Jun 2016 13:40:29 GMT

Try the following:

index=* | timechart span=1m dc(Host_Name) as Count_Of_Hosts 
| streamstats window=2 last(Count_Of_Hosts) AS Last, first(Count_Of_Hosts) AS First 
| eval Delta=Last-First

So, we have our base search then feed that to timechart splitting things by month. You can TOTALLY use bin+stats, which would be more efficient, but I think as an example this is more clear yet is only slightly slower.

That goes into streamstats which takes those events two at a time ( window=2 ) and sets Last and First to the count from the existing event and the preceding event. This makes the last line a simple subtraction on each event.

As it stands, all those fields will be output, so you can see what it does yourself and confirm. If you don't need those extra fields, you can always | table ... the ones you want or | fields - ... to get rid of the ones you don't want.

For reference in case I fat fingered the changes to fit your own use case better, here's the test I ran on my laptop (which has only local logs and miscellaneous stuff I'm testing for answers):

index=* | timechart span=1w dc(eventtype) as EventTypes | streamstats window=2 last(EventTypes) AS Last, first(EventTypes) AS First | eval Delta=Last-First

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

sundareshr — Fri, 10 Jun 2016 13:42:09 GMT

Try this

index=* earliest=-1mon@mon | eval when=if(_time<relative_time(now(), "@mon"), "Prev", "Current")  | chart dc(host) as h over host by when | eval action=case(Current>Prev, "Added", Current<Prev, "Removed", 1=1, "No Change") | stats count by action

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Fri, 10 Jun 2016 13:58:00 GMT

This seems to be getting there!

| timechart span=1m dc(Host_Name) as Count_Of_Hosts 
 | streamstats window=2 last(Count_Of_Hosts) AS Last, first(Count_Of_Hosts) AS First 
 | eval Delta=Last-First

I'm seeing this in the UI

The specified span would result in too many (>50000) rows.

Could be faster without span and timechart? Maybe just stats?
Thanks!

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Fri, 10 Jun 2016 14:53:31 GMT

Also if I add 1 host and remove another host in a month, the stats will be the same and the delta zero but we had movement...

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Sun, 12 Jun 2016 21:03:47 GMT

Hi Rich,
What do you think? Is there away to show that movement of host?

Cheers

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

woodcock — Mon, 13 Jun 2016 20:12:46 GMT

do you just need the delta between the columns or do you need an accurate number? If you need accuracy, then you need to track the distinct values of Host_Name kept for each column and you need to compare these sets. The former is pretty easy, the latter, depending on the order of magnitude of your sets, much less so.

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Mon, 13 Jun 2016 20:16:54 GMT

an accurate number so movement + or - is tracked.

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

woodcock — Mon, 13 Jun 2016 22:05:36 GMT

First of all, DO NOT use the "free" (but wrong) date_* values; calculate your own or use | bucket _time span=1mon.

You can only calculate changes in distinctness by comparing each entire list against other lists (not by subtracting dc values) so I would use a monthly Summary Index to store a list of distinct hosts and then build your search to pull from that aggregate store. You can use sistats to help you with this:

http://docs.splunk.com/Documentation/Splunk/6.0.4/SearchReference/Sistats

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Tue, 29 Sep 2020 09:58:50 GMT

Hi Greg,
Still struggling with this can you provide anymore instructions? The data is summarized but im struggling with the query.

| bucket _time span=1mon | sistats dc(Host_Name) by _time

Thanks

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

woodcock — Tue, 21 Jun 2016 11:55:25 GMT

That should be pretty much it. What problem are you having, exactly? What is your SI-populating search?

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

smudge797 — Mon, 27 Jun 2016 12:27:44 GMT

Here is the SI search:

... source=*.csv | eval IMT = substr(Cost,2)| convert rmcomma(IMT) | eval Total=round(IMT,2) | stats sum("Total") as Cost by Date "Cost Center" "Org L5" "Org L6" Org_Description IIGL1 IIGL2 IIGL3 IIGL4 Product Feed_Name

Re: How to search the count of adds/removes (new hosts vs host decoms) month on month?

woodcock — Mon, 27 Jun 2016 12:42:46 GMT

I do not see sistats in your search and this search looks nothing like your original search.. If you are planning to us the dc function coming back out of your SI, you really need to use sistats (like I said in my answer) because it does all the crazy magic for you:

http://docs.splunk.com/Documentation/Splunk/6.4.1/SearchReference/sistats

 index=* | sistats dc(Host_Name) by date_month 

index=YourSummaryIndex | stats dc(Host_Name) by date_month

That is the answer for THIS question. However it appears that your needs have morphed completely away from the original question so perhaps you should start another question.