Clojure collector creates new network_userid for every event


#1

Our Snowplow installation has some sites using the Cloudfront collector, and some using an old version of Clojure (clj-0.6.0-tom-0.0.4).

We have never actually used network_userid in any analytics, so in the course of routine maintenance I ran some queries to see how many network_userid values were repeated across domains.

The surprise was that, not only are network_userids not showing up across domains, they are unique in atomic.events. That is, we are getting a fresh network_userid on every single event, even within domains.:

events=> select count(*), count(distinct event_id) event_ids, count(distinct page_urlhost) hosts, count(distinct domain_userid) domain_userids, count(distinct network_userid) network_userids from atomic.events                                                              
where v_collector = 'clj-0.6.0-tom-0.0.4' and collector_tstamp > '2018-03-01';
  count  | event_ids | hosts | domain_userids | network_userids 
---------+-----------+-------+----------------+-----------------
 4587548 |   4587225 |   113 |         431559 |         4587545

I know many browsers are blocking 3rd party cookies nowadays, but this seems extreme.

So, questions:

  • Is our clojure collector misconfigured? Or maybe our JS tracker code?
  • Would results be different if we updated to the current Clojure collector?
  • More broadly, given that 3rd party cookies are in decline, should we just bail on clojure and use the Cloudfront collector exclusively?

Any advice appreciated!


#2

Provided browser blocks 3rd party cookies, every new page viewed will generate new network_userid. This is normal :wink:

I would not change “active” collector to logs again. Especially if you see development being pushed forward into real-time pipeline direction.


#3

I understand about browsers blocking 3rd party cookies, but I didn’t expect 100 percent of browsers to be so strict; especially when I have enabled them on my own browser.

I looked at network cookie traffic in Chrome Developer Tools and found the culprit: Our tracker was set up to use a CNAME for the Clojure collector, and the collector was setting the “sp” cookie on its primary hostname.