Does dynatrace support APM capabilities for side car containers based on C++11 code base such as Envoy?

[gaurav_abbi]

I have a K8S cluster running on GKE. I have installed dynatrace one agent via the operator as mentioned in the documentation. Then I have installed the open source project Kyma. Now Kyma comes with Istio installed and sidecar injection enabled by default. What this implies is that any K8S deployment to this cluster will end up having Pod with two containers

1. Application container containing the domain logic
2. Istio-proxy which is a sidecar container that sits in front of the application containers and handles networking and other concerns around being a service mesh.

In the dynatrace environment UI, I can see the Pod listed as a process but it only shows one container in the details. I suspect this being due to the istio-proxy/envoy being based on C++11 codebase.

Additionally, what are the plans to enable pluggable tracing capabilities via Istio/envoy tracing vs doing the tracing by low-level code instrumentation? This will be useful for projects like Kyma which already use the sidecar pattern to achieve tracing and other monitoring concerns.

[skrystosik]

I think you will need manually instrument your native code according to this:

https://github.com/Dynatrace/OneAgent-SDK-for-C

Sebastian

[sonja · Accepted solution]

UPDATE: actually, since OneAgent version 180, we should automatically detect Envoy processes out-of-the-box. No custom process group detection rules required.

Hi @gaurav a. - we don't detect all envoy processes (yet) automatically out-of-the-box. You can configure a custom process group detection rules to start monitoring it.

E.g. this is how I have configured Dynatrace for my Istio environment (you would have to check if the executable name and path are the same for you):

As with most proxies, tracing through Envoy already works out-of-the-box. You should already be able to see end-to-end tracing.

But yes, we are already working on adding more visibility into Envoy - e.g. to be able to see how much time is spent into the sidecar proxies for each request.

Sonja