TL;DR
Pod 1 - sends errors to the K8S Cluster
Pod 2 - consumes errors from K8S Cluster and sends them to Sentry.io
Setup

Overview

1. This demo involves running 2 kubernetes pods and the kubernetes cluster.

2. Pod 1 requests too many resources (CPU/RAM) from the K8S cluster, and thus causes an error that gets sent into the Kubernetes Steam/log of Events

3. Pod 2 uses the Sentry Python SDK to capture this error from the Kubernetes Stream and send it as a Event to Sentry.io

4. Docker Images and yaml's are provided for both of these pods.

The docker image used for pod2 is here hub.docker getsentry/sentry-kubernetes.

The docker container runs sentry-kubernetes.py which watches/consumes the Kubernetes Stream and has logic for evaluating if the activity is something benign/ignorable or errorneous/problematic, in which case it uses Sentry's capture_message method to capture and send it to Sentry.io.

It decides for you what is an Event or not, but you could amplify the logic with your own by updating sentry-python.py and DIY re-making your own image/k8pod. This way, you can decide what constitutes an 'error' or anything you want to be notified about on Sentry.io. The source code is here getsentry/sentry-kubernetes/sentry-kubernetes.py

Some commands need to be run for serviceaccounts and clusterroles, or else the sentry-kubernetes POD will lack permission to reading from the Kubernetes cluster Stream and the stream will be full of Forbidden403's.

• Don't let the power of Capture Message stop here. You can use it in any major language or framework as it's available in all of our SDK's.

• You can any string piece of data as an input to sentry's capture_message function for sending it to Sentry.io as an Event. You can still use Sentry scope for adding Tags, Breadcrumbs and the Extra Information as well. Using Capture Message in key places provides more context to your overall tech stack, in addition to the usual places where you use Sentry (REST API's, front-end Javascript). This demo exemplifies this power/visibility by using it with Kubernetes.

• Do not use Capture Message as a complete replacement for any logging solutions in place. Reserve it for actionable information or for capturing something that might be otherwise difficult to hunt down and find in your GKE or existing logging solution.

kubectl: 1.16
kubectl client: v1.16.2 / kubectl server: v1.16.0
minikube version: v1.4.0

1. git clone [email protected]:sentry-native/sentry-native.git
2. create a project on Sentry.io and get your DSN (API) key ready for running the sentry-kubernetes container
3. start minikube
4. configure service accounts and clusterroles
5. start pods (2)
6. See event on Sentry.io

https://kubernetes.io/docs/setup/learning-environment/minikube/#quickstart

minikube start

kubectl create sa sentry-kubernetes
kubectl create clusterrole sentry-kubernetes --verb=get,list,watch --resource=events
kubectl create clusterrolebinding sentry-kubernetes --clusterrole=sentry-kubernetes --serviceaccount=default:sentry-kubernetes
or
kubectl create clusterrolebinding sentry-kubernetes --clusterrole=sentry-kubernetes --serviceaccount=sentry-kubernetes

kubectl run sentry-kubernetes \
  --image getsentry/sentry-kubernetes \
  --serviceaccount=sentry-kubernetes \
  --env="DSN=https://[email protected]/1783432"

# Specifies a Memory/CPU resource request that is too big for your nodes"  
kubectl apply -f ./cpu-request-limit-2.yaml
kubectl apply -f ./memory-request-limit-3.yaml

kubectl delete deployment hello-minikube
kubectl delete deployment sentry-kubernetes
kubectl delete services hello-minikub
kubectl delete -n default pod memory-demo-3
kubectl delete -n default pod cpu-demo-2

minikube stop
minikube delete

minikube -help

# launches http://127.0.0.1:50159/api/v1/namespaces/kubernetes-dashboard/services/http:kubernetes-dashboard:/proxy/#/overview?namespace=default
minikube dashboard
minikube addons list

kubectl get -h

kubectl get deployments
kubectl get pods
kubectl get pod sentry-kubernetes-5dbfb4597f-xr7kj
kubectl describe pod # lists all, with statuses
kubectl describe pod <name>
kubectl get events <-- see events that sentry-kubernetes isn't capturing
kubectl get services
kubectl config view
kubectl get namespaces
kubectl logs <name> <--- to see logs of the pod, not to be confused with Events of the pod.
kubectl get sa <---get service accounts
kubectl logs <sentry-kubernetes>
# shows info, look for sa/security/roles/clusterRolebinding
kubectl get deployment sentry-kubernetes -o yaml  

serviceaccounts and clusterroles

kb get pod sentry-kubernetes-69f9bbdfc7-mfn52 -o yaml
kb get clusterrole sentry-kubernetes -o yaml
kb get clusterrolebinding sentry-kubernetes -o yaml
and
kb get serviceaccounts/sentry-kubernetes -o yaml

troubleshooting...

kubectl scale deployment sentry-kubernetes --replicas=1
kubectl scale deployment sentry-kubernetes --replicas=0
kubectl create clusterrolebinding sentry-kubernetes --clusterrole=sentry-kubernetes --serviceaccount=sentry-kubernetes
kubectl delete clusterrolebinding sentry-kubernetes
kubectl create clusterrolebinding sentry-kubernetes --clusterrole=sentry-kubernetes --serviceaccount=default:sentry-kubernetes
kubectl create clusterrolebinding sentry-kubernetes --clusterrole=sentry-kubernetes --serviceaccount=sentry-kubernetes

Cheatsheet
https://kubernetes.io/docs/reference/kubectl/cheatsheet/

"Create a Pod that Requests too many resources from a Node" CPU or Memory
https://kubernetes.io/docs/tasks/configure-pod-container/assign-cpu-resource/#specify-a-cpu-request-that-is-too-big-for-your-nodes
https://kubernetes.io/docs/tasks/configure-pod-container/assign-memory-resource/

Cluster administration logging
https://kubernetes.io/docs/concepts/cluster-administration/logging/

Bonus Sentry Helm Charts
https://github.com/helm/charts/tree/master/stable/sentry

Bonus Other developers in the open-source community have made their own versions of this:
https://hub.docker.com/search?q=sentry-kubernetes&type=image
https://github.com/stevelacy/go-sentry-kubernetes

https://blog.sentry.io/2019/04/17/surface-kubernetes-errors-with-sentry/