Splunk Search

what is the difference between "... | timechart count by host" and "... | timechart span=1d count by host" ?

Explorer

The two queries:
search sourcetype="access" host="www" | timechart count by host
and
search sourcetype="access" host="www" | timechart span=1d count by host

The former query is an example from the tutorial claiming to yield a timechart of the hits on servers from a webfarm, but the hitcounts are plain false. It states that www1 received only 10 hit on the first day, and scarcely more on the following.
My modification shown in the second example produces the correct numbers (e.g. 394 hits for www1 on the first day).

What is it that "count" counts in the first example, and why is this so?
Any explanations?

Motivator

Hi phuehne

This request :

 ... | timechart count by host   

counts the number of events by host by time , here you can set you timerange like you want

This other request :

  .....| timechart span=1d count by host

counts the number of events by host every 1 day.

Thanks

0 Karma

Explorer

Okay, but if I don't set a timerange explicitely, what would the default be?

0 Karma

Motivator

Your two queries are identical with one caveat: The first one automatically sizes buckets based on the timeframe you're searching. (So if you search over an hour, it may bucket per minute. But if you change your search range to a week, it'll bucket by day) This means you can change your time range, and the chart will always be readable without having to modify the underlying search.

When you add the span= condition, you're explicitly defining how it groups the results. So if your search timeframe is a week, and you say span=1m then you'll get a record for each minute. (making for a very busy chart)

Again, other than that, the two searches are identical. They're producing counts of events by host over automatically or manually defined time buckets.

Explorer

Hello emiller,

I'm very puzzled,
I've installed my own splunk (version 6.2.2) on debian in the meantime and loaded the tutorial data into it according to the instruction in the tutorrial. But when I click on "Start to search", the reuslt is an orange triangle with ! in it and the messages "unknown sid" and "The search job XXX was canceled remotely or expired."
Searching for these messages on the web didn't yield any usable results. Also, hitting the "Data summary" button results in no data summary being displayed and a "waiting for results...".
So I returned to the sandbox to see for my former count problem there again. And ... mysterioulsy .. it was gone. I pasted the very same command "index=tutdata sourcetype="access" host="www" | timechart count by host" into the search field, and today it produced the correct values in the count column.
I didn't produce any screen shots last week, so I have no proof of the wrong results I wrote about.

Regards
Peter

0 Karma

Explorer

Thnaks for the prompt answer. Nevertheless, I didn't modify the timeframe between the two queries. It has always been set to "All times", which is one week for the tutorial data. In both cases, the timechatr's x-axis showed two days intervals - only the count values differed between the two query results.
The flexibility feature you explained in your answer would be very impressive, but still the chart reads:
"www1 had 10 hits on 04-03-2015",
and not
"www1 hat 10 hits between 20:00 and 21:00 of 04-03-2015".
This deems me quite missleading.

0 Karma

Splunk Employee
Splunk Employee

Which version of Splunk are each of you running?

0 Karma

Motivator

I'm on the newest version, personally.

0 Karma

Motivator

I am not seeing any discrepancy when I run the two searches against the test data.

Without explicit span

With explicit span

0 Karma