TIPS & TRICKS

analytics.usa.gov Recreated Using Splunk

analytics-gov

Have you guys seen analytics.usa.gov?

It’s a great break-down of web traffic to various US government sites. I’m a passionate believer in the open-data movement, and this is simply wonderful! A very big pat on the back to the US Government from across the other-side of the Atlantic. I’m now looking at you now Europe…

What’s more, the team that built the app have exposed API endpoints for the data that’s currently being displayed. Which – to my excitement – allows us to start playing with it in Splunk.

I wanted to show you just how easy it is to recreate the site in Splunk, and why you would want to do it in the first place.

Why Reinvent the Wheel?

I hear you. Why fix something that’s not broken?

Data often becomes much more interesting when you can start to compare it to other datasets and make correlations. One of the things Splunk does well is to easily bring all types of machine data into one place.

For example, if it’s that time of the year we all love when taxes are due to be filed; we could examine the impact it has on the volume of traffic to certain parts of the IRS site.

Also, although near-real time updates are great, using Splunk we can also store the data locally to get a historic view of the data to identify any trends.

How Can I Do It?

Set Up A REST Input in Splunk

To poll the API endpoints on analytics.usa.gov we can install the REST API Modular Input. Grab it here.

You can see a list of the API endpoints available on http://analytics.usa.gov under “download the data”.

Once the endpoints you want have been identified, in the Splunk UI navigate to:

Settings > Data Inputs > REST > Add New

At a very minimum you need to define the fields:

Name: <what ever you like>
Endpoint URL: <e.g: https://analytics.usa.gov/data/live/realtime.json>
HTTP Method: GET
Authentication Type: none
Polling Interval: <set this to be identical the value the endpoint is updated. e.g if once per minute, set this value as 60 seconds.
Sourcetype: Manual (dropdown option) > _JSON

Cheat code: To make things quicker, just paste the text in this Gist to inputs.conf (in the location below). This way all the inputs currently available will be automatically configured for you.

$SPLK/etc/apps/launcher/local/inputs.conf

Note, you may need to create the directory and inputs.conf file.

Build A Dashboard Using Search

Wait… 2 things to note before doing this,

  • these searches will only work if the “sourcetype” and “source” you’ve used are the same as mine, otherwise adjust them as required (if you copy the Gist linked above you’ll be fine).
  • to save a search as a dashboard visualisation, first run the search, and then click “Save As” in the top right hand corner of the page.

Open up the Splunk Search Bar and enter the following queries:

Visitors online now:

sourcetype="_json" source="rest://people online right now 1min"| stats latest("data{}.active_visitors") as "Active Visitors"

Visitors in the past 90 days:

sourcetype="_json" source="rest://visits by desktop mobile tablet devices over 90 days 1day" | stats latest(totals.visits) AS "Total Visits"

Visitor device types in the past 90 days:

sourcetype="_json" source="rest://visits by desktop mobile tablet devices over 90 days 1day" | stats latest(totals.devices.desktop) AS "Desktop Visitors" latest(totals.devices.mobile) AS "Mobile Visitors" latest(totals.devices.tablet) AS "Tablet Visitors" | transpose | rename "column" AS "Device Type" "row 1" AS "Visits" | sort -"Visits"

Visitor OS types in the past 90 days:

sourcetype="_json" source="rest://visits broken down by operating system over 90 days 1day" | stats latest("totals.os.Android") AS "Android" latest("totals.os.BlackBerry") AS "Blackberry" latest("totals.os.iOS") AS "iOS" latest("totals.os.Linux") AS "Linux" latest("totals.os.Macintosh") AS "Macintosh" latest("totals.os.Other") AS "Not Listed" latest("totals.os.Windows") AS "Windows" latest("totals.os.Windows Phone") AS "Windows Phone" | transpose | rename "column" AS "Operating System" "row 1" AS "Visitors" | sort -"Visitors"

Visitor Windows version in the past 90 days:

sourcetype="_json" source="rest://visits broken down by Windows version over 90 days 1day" | stats latest("totals.os_version.7") AS "7" latest("totals.os_version.8") AS "8" latest("totals.os_version.8.1") AS "8.1" latest("totals.os_version.Other") AS "Not Listed" latest("totals.os_version.Vista") AS "Vista" latest("totals.os_version.XP") AS "Vista" | transpose | rename "column" AS "Windows Version" "row 1" AS "Visitors" | sort -"Visitors"

Visitor browser types in the past 90 days:

sourcetype="_json" source="rest://visits broken down by browser over 90 days 1 day" |  stats latest("totals.browser.Amazon Silk") AS "Amazon Silk" latest("totals.browser.Android Browser") AS "Android Browser" latest("totals.browser.BlackBerry") AS "Blackberry" latest("totals.browser.Chrome") AS "Chrome" latest("totals.browser.Firefox") AS "Firefox" latest("totals.browser.IE with Chrome Frame") AS "IE (Chrome Frame)" latest("totals.browser.Internet Explorer") AS "IE" latest("totals.browser.Opera") AS "Opera" latest("totals.browser.Opera Mini") AS "Opera Mini" latest("totals.browser.Other") AS "Not listed" latest("totals.browser.Safari") AS "Safari" latest("totals.browser.Safari (in-app)") AS "Safari (in-app)" latest("totals.browser.UC Browser") AS "UC Browser" | transpose | rename "column" AS "Browser" "row 1" AS "Visits" | sort -"Visits"

Visitor IE Version in the past 90 days:

sourcetype="_json" source="rest://visits broken down by Internet Explorer version over 90 days 1day" | stats latest("totals.ie_version.10.0") AS "10.0" latest("totals.ie_version.11.0") AS "11.0" latest("totals.ie_version.6.0") AS "6.0" latest("totals.ie_version.7.0") AS "7.0" latest("totals.ie_version.8.0") AS "8.0" latest("totals.ie_version.9.0") AS "9.0" latest("totals.ie_version.Other") AS "Not Listed" | transpose | rename "column" AS "IE Version" "row 1" AS "Visitors" | sort -"Visitors"

Cheat code: you can copy the XML in this Gist and paste it to create a prebuilt dashboard containing the searches above.

----------------------------------------------------
Thanks!
David Greenwood

Splunk
Posted by

Splunk