If you wish you can setup your Open-OnDemand instance to send usage data to Google Analytics (GA) that you can then query and report on, this page walks through how to do just that.

Note

You’ll need to have a Google Analytics account setup as a prerequisite to this.

To query GA You’ll need to have to have a service account setup with the appropriate permissions. See info on Google service accounts and Google IAM roles for more details on that.

## Configure Open OnDemand¶

# /etc/ood/config/ood_portal.yml
---

analytics:
# the id will be specific to your account, but url is likely the same
id: UA-99999999-1


This configuration will generate a block similar to this in your apache’s ood-portal.conf file (after running the ood-portal-generator).

<Location "/pun">
...

SetEnv OOD_ANALYTICS_TRACKING_ID  "UA-99999999-1"
LuaHookLog analytics.lua analytics_handler
</Location>


You’ll need to create all of these custom dimensions and custom metrics in the appropriate GA account(s).

Warning

Order matters here! Index numbers are given to ensure you create and define these items in the correct order. Otherwise Google Analytics will be incorrectly indexing these metrics.

As an example say Username gets index 3 instead of index 1. Now when you query for dimension3 thinking it’s timestamps, you’ll get back usernames instead!

Table 1.1 GA custom dimensions
Name Index Scope
Session ID 2 Session
Timestamp 3 Hit
Request Method 5 Hit
Request Status 6 Hit
Document Referrer 7 Hit
Table 1.2 GA custom metrics
Name Index Scope Formatting Type
Proxy Time 1 Hit Integer
User Map Time 2 Hit Integer

Now that you have Open-OnDemand sending information to GA and it’s all configured correctly, you can now query GA for this information, parse it and present it in any fashion you like.

Here’s a small portion of how we query GA in ruby, but there are many GA client libraries available.

This example is not complete and is only meant to illustrate how to query GA given the defined metric set above. Let’s go through each of these things.

# Dimensions - here we want dimensions 1, 3 and something called pagePath which is the web
# page requested. pagePath is a google predefined dimension that we populated. Dimensions 1
# and 3 were created above and are the username and timestamp (this is why the order in
# which they're defined is important).
DIMENSIONS = %w(
ga:dimension3
ga:dimension1
ga:pagePath
)

# we only want to report the hit metrics
METRICS    = %w(
ga:hits
)

# First we specify the host so that we only get metrics from a specific host. Secondly,
# we filter only only 200 responses (dimension6 is status code) and we don't want to
# report on file editor edits.
FILTERS    = %W(
ga:hostname==#{HOST};ga:dimension6==200;ga:pagePath!=/pun/sys/file-editor/edit
)

# now we can create our analytics object and make the query

# Here we query for the data that we want. A lot of things are omitted in this example
# for brevity like START_DATE (dynamic query times like the first day of the month)
# or GA_PROFILE (part of our credentials). And the fact that this is in a loop paginating
# the results, updating 'start_index' and only requesting STEP_SIZE (10,000 in our case)
# results at a time.
results = analytics.get_ga_data(
"ga:#{GA_PROFILE}",
START_DATE,
END_DATE,
METRICS.join(','),
dimensions:  DIMENSIONS.empty? ? nil : DIMENSIONS.join(','),
filters:     FILTERS.empty?    ? nil : FILTERS.join(','),
sort:        SORT.empty?       ? nil : SORT.join(','),
start_index: start_index,
max_results: STEP_SIZE
)

target = open('my-report', "w")

# now we can write out the results in a format that I want for my reporting.
results.rows.each do |row|
begin
app = row[2]
row[2] = parse_uri(app, user: row[1])
row << app
target.write "#{row.join('|')}\n"
end