Understanding Azure DevOps Pipeline Statistics

Pipeline statistics in Azure DevOps provide a wealth of information that can help you understand the performance and efficiency of your pipelines. Even with really basic informations you can have interesting information. Lets examine

Figure 1: Pipeline statistics dashboard in Azure DevOps

My favorite information (1) is the 80th percentile of pipeline duration, this indicates me if for some reason some pipeline is becoming slower. Actually in Figure 1 I can see that seems to be an increase in pipeline duration. With (2) you can immediately understand which is the step that predates the execution time. In this example it is not a mistery, the IN-PROCESS step is actually running a full suite of integration test that uses MongoDB, Elasticsearch, and a complex software end-to-end without the UI.

Clicking on the tiles brings you to a detailed page that shows much more information about duration.

Figure 2: Detailed statistics about build time

As you can see I have the ability to have a graph that show the trend, up to a 180 days of data. In this example I see no mistery, we spend time running test, then building the .NET solution (94 projects, 12 years of code) then we have again jest tests and so on. Nevertheless it is interesting looking at these graph because sometimes you find something that can surprise you.

You can choose the number of days but also you can filter by branch. Usually we are interested only in develop branch.

If you look again at Figure 1 the central graph is devoted to test failures, in this situation the detail is the real information you want to look at.

Figure 3: Detailed statistics about test failures

As you can see we have a high number of failure today, this happens due to my fault because I’ve added test on ElasticSearch 8, forgetting to configure a valid ElasticSearch 8 instances in build server. Actually the tests should be skipped, but a bug in the logic made the test run and they all fails because no ElasticSearch is found.

You can filter by branch, by test status, but the most important thing is that you can click on a test failure to look at details of that specific tests.

Figure 4: Detailed statistics about single test failure

You have also the list of all the branches where the test is failing, this helps you to determine if it is failing on develop or master branch or it fails on a feature branch that still needs to be merged with the stable branches.

To conclude: always keeps an eye on the basic metrics of your pipeline, you will have a really good insight of how your project is going.

Gian Maria.