What happens when the master of Coubernes fails?

I tried to figure out what happens when the Kubernetes wizard fails in a cluster that has only one wizard. Do web requests route to containers if this happens, or is the entire system just shutting down?

According to the OpenShift 3 documentation, which is built on top of Kubernetes, ( https://docs.openshift.com/enterprise/3.2/architecture/infrastructure_components/kubernetes_infrastructure.html ), if the wizard does not work, the nodes continue to function properly, but the system loses the ability to control the pods. Is it the same for vanilla kubernetes?

+8
kubernetes openshift-origin
source share
1 answer

I understand that the wizard launches the API, and now (starting from version 1.3?) It manages the underlying cloud infrastructure. When it is offline, the API will be disabled, so the cluster will cease to be a cluster and will instead become a bunch of ad-hoc nodes over this period. The cluster will not be able to respond to node errors, create new resources, move containers to new nodes, etc. Until the wizard is connected to the network.

However, in any case, life for applications will continue as usual, if during this time there is no significant failure, * or nodes are rebooted, because TCP / UDP services, load balancers, DNS, dashboard, etc. If everyone continues to function.

After some reading, it appears that DNS queries may not resolve correctly if node reboots at this time. I definitely recommend that you do not restart nodes during this period if you are concerned that your services remain available.

If you want to test this yourself, there is a handy utility called minikube that you can use to "try before you buy." Remember (if you're from the future) that Kubernetes is still changing fast, and this answer may not be true anymore while you are.

+14
source share

All Articles