Building a Kubernetes Operator for Automated Distributed Tracing - Orr Weinstein, Lumigo

Building a Kubernetes Operator for Automated Distributed Tracing - Orr Weinstein, Lumigo

Implementing distributed tracing–-and all the required instrumentation–hasn’t been for the faint-hearted. But in this talk, we present a Kubernetes operator that allows you to go from no distributed tracing in your Kubernetes cluster, to tracing every javascript, python and Java pod in a namespace, by creating only one Kubernetes secret and one custom resource.

Some of the topics we will be covering are:
- Admission controller mutating webhooks
- Creating events on injected resources
- Reconciling events created by the webhook

[jonathanlim]

That’s nice. In the case of using a mutatingwebhookconfiguration, how do you ensure that in the event of a node restart, and when all workloads are rescheduled, that the operator will get prioritized to spin up before the application pods?

If I understood it correctly, in the scenario where application pod spins up before operator, then those app pods won’t be injected with the tracers.