Today we have metrics, which provides information like how often a particular event occurs or the quantity of a resource being consumed. This is great for things like tracking the how long it takes to respond to requests, how much CPU is consumed, etc.

What we are missing is deeper insight into the application. To continue with the example we may know that on average requests are serviced around 350ms. What we don't know is why. Is it a long database call? Is an underlying service slow to respond? Is there a blocking operation causing resource contention?

These questions can be answered with Tracing, which provides an in-depth look at a particular request, tracing its flow from initial reception to response to the service.

Here is an example of the insight you can glean with tracing:

Proposed Iterations:

1. Link to Jaeger: https://gitlab.com/gitlab-org/gitlab-ee/issues/4753
2. Deploy ElasticSearch to cluster: https://gitlab.com/gitlab-org/gitlab-ee/issues/5694
3. Jaeger Authentication: https://gitlab.com/gitlab-org/gitlab-ee/issues/8498
4. Deploy Jaeger to cluster: https://gitlab.com/gitlab-org/gitlab-ee/issues/5182