google analytics in nginx

nginx_google_analytics

# send logs for all traffic (including non-html) to google analytics
#
# in server block:
#  set $google_analytics_id "UA-THECORRECT-ID";
#  include            /srv/nginx/google_analytics;
#
# in location blocks:
#  post_action @ga;
#
# notes: post_action has been referred to by nginx devs as a "dirty hack" and this approach may stop working one day.
#        if so, can fall back to the js injection method
#
# refs:
# https://developers.google.com/analytics/devguides/collection/protocol/v1/
# http://developers-club.com/posts/260553/
# https://qzaidi.github.io/2014/04/19/google-universal-analytics/

userid on;
userid_name cid;
userid_expires 2y;

set $ga_dp $uri;
set $ga_host $host;

#userid is contained in uid_got or uid_set, depending on if it's already been set
#also the cookie variable name is included (e.g. cid=ASDF), so no need to specify it in the query params
set $cid $uid_got$uid_set;

location @ga {
  internal;

  #don't track web asset file requests
  if ($request_filename ~* .(gif|jpe?g|png|ico|js|css)$ ) {
    return 200;
    break;
  }

  #don't track robots.txt
  if ($request_filename ~* robots.txt$ ) {
    return 200;
    break;
  }
  
  #don't track internal requests
  #NOTE: this requires a geo section in the http block listing CIDRs for internal addresses,
  #usually in nginx.conf
  #example:
  #geo $internal_request {
  #  default 0;
  #    123.44.15.122/24 1;
  #    109.34.13.98/32 1;
  #}

  if ($internal_request) {
    return 200;
    break;
  }

  #NOTE: this does *not* currently try to exclude requests from bots.
  #known bots can be excluded from Google Analytics in the Admin settings under
  #View/View Settings/Exclude all hits from known bots and spiders
  #if its desirable to exclude specific user agents here, define a map:
  #
  #map $http_user_agent $bot_request {
  #   default 0;
  #   ~*(ahrefsbot|dotbot|semrush) 1;
  #   ~*(yandex|baidu) 1;
  #}
  #
  #then exclude these requests from being sent to Google Analytics
  #note: these requests would still be allowed, they just wouldn't be tracked in GA

  resolver 8.8.8.8 ipv6=off;
  proxy_ignore_client_abort on;
  proxy_next_upstream timeout;
  proxy_connect_timeout 1s;

  set $ga_api "https://www.google-analytics.com";
  #set $ga_api "https://www.google-analytics.com/debug"; #use for debugging

  #TESTING
  #Route request to a local target to debug the constructed Measurement API url
  #vanilla nginx will do:
  #docker run --rm -p 8888:80 nginx:1.12
  #or something more thorough:
  #docker run --rm -p 8888:80 mendhak/http-https-echo
  #set $ga_api "http://192.168.0.45:8888"; #for local testing

  #NOTE: DOWNLOAD SIZE TRACKING
  #Google Analytics doesn't track download sizes out of the box.
  #Here we are using a custom metric (cm1) and sending the value of nginx's body_bytes_sent
  #The custom metric has to be set up in Google Analytics (Admin/Property/Custom Definitions/Custom Metrics)
  #and then added to a custom report (Customization/Custom Reports) before it can be visualized in Google Analytics.

  proxy_set_header Host "www.google-analytics.com";
  proxy_pass $ga_api/collect?v=1&dp=/$ga_host$ga_dp&dh=$ga_host&tid=$google_analytics_id&$cid&uip=$remote_addr&t=pageview&cm1=$body_bytes_sent;
}

nginx_google_analytics_example_usage

server {
  listen              80;
  server_name         erddap.yoursite.org;

  #set $google_analytics_id and include the google_analytics file below
  set $google_analytics_id "UA-12345678-9";
  include            /etc/nginx/google_analytics;

  location /erddap/ {
    #your normal config goes here...

    #call the @ga location as a post_action
    post_action        @ga;
  }
}