The master node serves as a kernel component in the Kubernetes system. Its duties include the following:
- Pushing and pulling information from etcd servers
- Acting as the portal for requests
- Assigning tasks to nodes
- Monitoring the running tasks
Three major daemons enable the master to fulfill the preceding duties; the following diagram indicates the activities of the aforementioned bullet points:
As you can see, the master is the communicator between workers and clients. Therefore, it will be a problem if the master crashes. A multiple-master Kubernetes system is not only fault tolerant, but also workload-balanced. It would not be an issue if one of them crashed, since other masters would still handle the jobs. We call this infrastructure design high availability, abbreviated to HA. In order to support HA structures, there will no longer be only one API server for accessing datastores and handling requests. Several API servers in separated master nodes would help to solve tasks simultaneously and shorten the response time.