You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After the upstream change apache/openwhisk#4041, the DockerContainerFactory in Kubernetes started failing in TravisCI . In particular, the /invokers/healthy/count route on the controller always returns 0 because the invoker never transitions from Unhealthy to Healthy.
The problem appears to be that log collection of the InvokerHealth action fails, resulting in the container being removed without a response being sent back to the controller. The failure with log collection may be limited to kubeadm-dind clusters on ubuntu 16.04 (log collection does not fail using Kubernetes on Docker for Mac).
This is likely an existing bug that was masked because before 4041, an activeack for the HealthAction was returned before log collection failed.
The text was updated successfully, but these errors were encountered:
Yep, specific to docker-in-docker kubernetes clusters. We need to mount /dind/docker/containers instead of /var/lib/docker/containers in the invoker pod.
After the upstream change apache/openwhisk#4041, the DockerContainerFactory in Kubernetes started failing in TravisCI . In particular, the
/invokers/healthy/count
route on the controller always returns 0 because the invoker never transitions from Unhealthy to Healthy.The problem appears to be that log collection of the InvokerHealth action fails, resulting in the container being removed without a response being sent back to the controller. The failure with log collection may be limited to kubeadm-dind clusters on ubuntu 16.04 (log collection does not fail using Kubernetes on Docker for Mac).
This is likely an existing bug that was masked because before 4041, an activeack for the HealthAction was returned before log collection failed.
The text was updated successfully, but these errors were encountered: