Configured my yml with failure accrual circuit breaking, with below config as reference. https://github.com/linkerd/linkerd-examples/blob/master/failure-accrual/linkerd.yml
client:
failureAccrual:
kind: io.l5d.consecutiveFailures
failures: 5
backoff:
kind: constant
ms: 10000
After applying the changes, I see the pods in crashloopbackoff state.
kubectl apply -f ./linkerd-fa.yml
I think you should delete the pods which is Crashloopbackoff and then run kubectl get pods
.
Crashloopbackoff
means that the process is failing to start. If you use kubectl logs pod/$POD_ID l5d
, you should see an error message. I'm guessing there's a formatting issue with your configuration that is causing linkerd to error out during startup.
Linkerd provides failure accrual by default as part of each of your client configurations. The default failure accrual configuration is 5 consecutive failures, which is also what you've pasted in the YAML snippet with your question. So in this case you could omit the failureAccrual setting altogether from your config and the behavior will be the same.