We have a kubernetes cluster with Rancher v1.21.10+rke2r2 and a local installation of dynatrace managed. We tried to install dynatrace full stack monitoring with the dynatrace operator as this is the installation which is recommended by dynatrace. Monitoring of kubernetes worked but not the monitoring of the applications in the cluster (Java, ga, webserver etc.).
The support answer is now, that dynatrace does not support kubernetes with rancher and rke, but it my work with other installation options.
Did someone succeed installing dynatrace operator and monitoring kubernetes with rancher and rke (monitoring kubernetes and the applications running in the pods)? Any hints, which konfigurations may work?
Thank you for any hints.
As I understand the post, the problem there is how to monitor rancher itself. We like to monitor the whole kubernetes cluster, which is installed with rancher and the rancher kubernetes engine. We want to monitor the applications running in pods in the cluster and the cluster itself. So I would say, it is not solved.
The dynatrace support answered, that rancher and rke is a not supported installation ("unsupported k8s distribution"), but 'there might be a workaround' using other installation options of dynatrace than using classic full stack with the dynatrace operator.
What you mean by this answer? We connected to a local Kubernetes API endpoint automatically as documented. Do we have to use the public Kubernetes API option for Ranger? See documentation
Hello @matthias_dillie ,
we have a customer running RKE2 and the Dynatrace Operator installation worked. There were only some issues with pulling the images, as RKE2 used proxy to pull them and avoided connecting to container image repository located on the Dynatrace Managed. There are only some issues with connecting the platform monitoring and the data from OneAgents as the OneAgent is not able to fetch some metadata from the containerd due to its location. However deep monitoring works without issues. Dynatrace Operator 0.5.0 and OA 1.237.
Thanks for your reply. Can you give us some more details on the installation? We tried classic full stack injection (https://www.dynatrace.com/support/help/shortlink/dto-deploy-options-k8s#classic) which is the recommended option, but this did not work.
@matthias_dillie can you please share what went wrong? Maybe share logs of the dynatrace operator and errors you are encountering. There are many points where a deployment into k8s may fail depending on your environment.
@Julius_Loman we first tried with classic full stack option. Monitoring Kubernetes itself worked good. The problem was, that we did not see the deep monitoring part of the applications running in the pods eg. java (OpenLiberty), go (Prometheus), (ingress-)nginx and so on. We saw the processes and cpu, memory and so they use but we did not see service calls, methods and all this details. With the dynatrace support, we looked at the system and found no errors in logfiles or other hints. The answer then was: not supported. See https://one.dynatrace.com/hc/en-us/requests/39972
Now we try with 'Automatic application-only injection'. We see now the OpenLiberty applications. This is ok now. But ingress-nginx, go and so on does not work. We no longer see this processes in Dynatrace. Maybe we have a ressource problem as we see a tmp.xxxx-file in the agent directory in the pods with go applications and ngnix instances?
@matthias_dillie I don't have access to your support tickets. Did support reveal any details on where is the injection failing with classic fullstack?